SQpeciual thaeme ntum - ERCIM News 112 · SQpeciual thaeme ntum Computing Also in this issue: Research and Innovation: Computers that negotiate on our behalf. ERCIM News is the magazine

ERCIM NEWSNumber 112 January 2018

www.ercim.eu

Special theme

QuantumComputing

Also in this issue:

Research and Innovation:

Computers that negotiate

on our behalf

ERCIM News is the magazine of ERCIM. Published quar-

terly, it reports on joint actions of the ERCIM partners, and

aims to reflect the contribution made by ERCIM to the

European Community in Information Technology and Applied

Mathematics. Through short articles and news items, it pro-

vides a forum for the exchange of information between the

institutes and also with the wider scientific community. This

issue has a circulation of about 6,000 printed copies and is

also available online.

ERCIM News is published by ERCIM EEIG

BP 93, F-06902 Sophia Antipolis Cedex, France

Tel: +33 4 9238 5010, E-mail: [email protected]

Director: Philipp Hoschka, ISSN 0926-4981

Contributions

Contributions should be submitted to the local editor of your

country

Copyright�notice

All authors, as identified in each article, retain copyright of their

work. ERCIM News is licensed under a Creative Commons

Attribution 4.0 International License (CC-BY).

Advertising

For current advertising rates and conditions, see

http://ercim-news.ercim.eu/ or contact [email protected]

ERCIM�News�online�edition��

http://ercim-news.ercim.eu/

Next�issue

October 2017, Special theme: Digital Humanities

Subscription

Subscribe to ERCIM News by sending an email to

[email protected] or by filling out the form at the

ERCIM News website: http://ercim-news.ercim.eu/

Editorial�Board:

Central editor:

Peter Kunz, ERCIM office ([email protected])

Local Editors:

Austria: Erwin Schoitsch ([email protected])

Cyprus: Georgia Kapitsaki ([email protected]

France: Steve Kremer ([email protected])

Germany: Michael Krapp ([email protected])

Greece: Lida Harami ([email protected]),

Artemios Voyiatzis ([email protected])

Hungary: Andras Benczur ([email protected])

Italy: Maurice ter Beek ([email protected])

Luxembourg: Thomas Tamisier ([email protected])

Norway: Poul Heegaard ([email protected])

Poland: Hung Son Nguyen ([email protected])

Portugal: José Borbinha, Technical University of Lisbon

([email protected])

Spain: Silvia Abrahão ([email protected])

Sweden: Kersti Hedman-Hammarström ([email protected])

Switzerland: Harry Rudin ([email protected])

The Netherlands: Annette Kik ([email protected])

W3C: Marie-Claire Forgue ([email protected])

Editorial Information Contents

ERCIM NEWS 112 January 2018

JoINt ERCIM ACtIoNS

4 Foreword from the President

4 Jos Baeten, Elected President of

ERCIM AISBL

5 Tim Baarslag Winner of the 2017

Cor Baayen Young Researcher

Award

6 ERCIM Cor Baayen Award 2017

7 ERCIM Established Working

Group on Blockchain Technology

7 HORIZON 2020 Project

Management

7 ERCIM “Alain Bensoussan” Fel-

lowship Programme

SPECIAL tHEME

The special theme section “QuantumComputing” has been coordinated byJop Briët (CWI) and Simon Perdrix(CNRS, LORIA)

8 Quantum Computation and

Information - Introduction to the

Special Theme

by Jop Briët (CWI) and SimonPerdrix (CNRS, LORIA)

10 How Classical Beings Can Test

Quantum Devices

by Stacey Jeffery (CWI)

11 Keeping Quantum Computers

Honest (or Verification of

Quantum Computing)

by Alexandru Gheorghiu(University of Edinburgh) andElham Kashefi (University ofEdinburgh, CNRS)

13 Experimental Requirements for

Quantum Computational

Supremacy by Boson Sampling

by Alex Neville and Chris Sparrow(University of Bristol)

14 From Classical to Quantum

Information – Or: When You

Have Less Than No Uncertainty

by Serge Fehr (CWI)

15 Preparing Ourselves for the

Threats of the Post-Quantum Era

by Thijs Veugen (TNO and CWI),Thomas Attema (TNO), Maran vanHeesch (TNO), and Léo Ducas(CWI)

17 Quantum Cryptography Beyond

Key Distribution

by Georgios M. Nikolopoulos(IESL-FORTH, Greece)

19 Quantum Lego: Graph States for

Quantum Computing and

Information Processing

by Damian Markham (LIP6, CNRS- Sorbonne Université)

ERCIM NEWS 112 January 2018 3

20 Graph Parameters and Physical

Correlations: from Shannon to

Connes, via Lovász and Tsirelson

by Simone Severini (UniversityCollege London)

22 High-Speed Entanglement

Sources for Photonic Quantum

Computers

by Fabian Laudenbach, SophieZeiger, Bernhard Schrenk andHannes Hübel (AIT)

23 A Formal Analysis of Quantum

Algorithms

by Benoît Valiron (LRI –CentraleSupelec, Univ. ParisSaclay)

24 Diagram Transformations Give a

New Handle on Quantum

Circuits and Foundations

by Aleks Kissinger (RadboudUniversity)

27 Quantum Computers and Their

Software Interfaces

by Peter Mueller, Andreas Fuhrerand Stefan Filipp (IBM Research –Zurich)

28 Chevalley-Warning Theorem in

Quantum Computing

by Gábor Ivanyos and Lajos Rónyai(MTA SZTAKI, Budapest)

30 Hardware Shortcuts for Robust

Quantum Computing

by Alain Sarlette (Inria) and PierreRouchon (MINES ParisTech)

31 Quantum Gates and Simulations

with Strongly Interacting

Rydberg Atoms

by David Petrosyan (IESL-FORTH)

33 Control of Quantum Systems by

Broken Adiabatic Paths

by Nicolas Augier (Écolepolytechnique), Ugo Boscain(CNRS) and Mario Sigalotti (Inria)

34 The Case for Quantum Software

by Harry Buhrman (CWI andQuSoft) and Floor van de Pavert(QuSoft)

RESEARCH ANd INNovAtIoN

This section features news aboutresearch activities and innovativedevelopments from Europeanresearch institutes

36 Computers that Negotiate on Our

Behalf

by Tim Baarslag (CWI)

38 Faster Text Similarity Using a

Linear-Complexity Relaxed Word

Mover’s Distance

by Kubilay Atasu, VasileiosVasileiadis, Michail Vlachos (IBMResearch – Zurich)

40 The CAPTCHA Samples Website

by Alessia Amelio (University ofCalabria), Darko Brodić, SanjaPetrovska (University of Belgrade),Radmila Janković (Serbian Academyof Sciences and Arts)

41 APOPSIS: A Web-based Platform

for the Analysis of Structured

Dialogues

by Elisjana Ymeralli, TheodorePatkos and Yannis Roussakis (ICS-FORTH)

42 Can we Trust Machine Learning

Results? Artificial Intelligence in

Safety-Critical Decision Support

by Katharina Holzinger (SBAResearch), Klaus Mak, (AustrianArmy) Peter Kieseberg (St. PöltenUniversity of Applied Sciences),Andreas Holzinger (MedicalUniversity Graz, Austria)

44 Formal Methods for the Railway

Sector

by Maurice ter Beek, AlessandroFantechi, Alessio Ferrari, StefaniaGnesi (ISTI-CNR, Italy), andRiccardo Scopigno (ISMB, Italy)

45 The Impacts of Low-Quality

Training Data on Information

Extraction from Clinical Reports

by Diego Marcheggiani (Universityof Amsterdam) and FabrizioSebastiani (CNR)

EvENtS, IN bRIEf

Reports46 4th International Indoor

Positioning and Indoor Navigation

Competition

47 Joint 22nd International

Workshop on Formal Methods for

Industrial Critical Systems and

17th International Workshop on

Automated Verification of Critical

Systems

by Ana Cavalcanti (University ofYork), Laure Petrucci (LIPN, CNRS& Université Paris 13) and CristinaSeceleanu (Mälardalen University)

48 10th International Conference of

the ERCIM Working Group on

Computational and

Methodological Statistics

Announcements48 Visual Heritage 2018

48 IFIP Networking 2018

48 Web Science Conference 2018

49 CIDOC Conference 2018

49 Postdoc Position in the Research

Project“Approximation

Algorithms, Quantum

Information and Semidefinite

Optimization” at CWI

50 ERCIM Membership

In Brief51 CWI hosts EIT Digital’s New

Innovation Space

51 CWI merges with NWO Institutes

Organisation

51 Google Awards Grant for fake

News detection to FORTH and

University of Cyprus


foreword

from the

President

ERCIM is a great organisa-tion with a lot of potential. Iam proud that I am the thirdCWI director sinceERCIM’s foundation in1989 who has become aPresident of ERCIM, after

Cor Baayen and Gerard van Oortmerssen. I will outline myideas on ERCIM’s strategy below.

The number of ERCIM Members has declined in recentyears. I want to focus more on our roots, what we are, andhow we started: as an association with a focus on researchinstitutes in the fields of both informatics and mathematics.Of course, leading research universities are also welcome tobecome ERCIM members, as we have university membersnow. Our focus will distinguish us from other organisations.

Further, experience has taught us that it is not good to limitthe number of members per country to one. This led to artifi-cial consortia that fell apart at some point. The model of oneorganisation or consortium per country has failed. We canand should have more ERCIM members per country. Forinstance, not only CWI could be a member in theNetherlands but also TNO. The Board of Directors shouldsee it as a joint effort to recruit new members.

I want to maintain ERCIM’s achievements, namely: • ERCIM News. This magazine is widely appreciated and is

becoming ever more important.• Awards. In addition to the Cor Baayen Award for the more

fundamentally orientated research, we could establish aninnovation award, which would be a good supplement andwhich is in line with the increased importance of valorisa-tion.

• Programs such as the ERCIM Alain Bensoussan Fellow-ship programme and the PhD program we are currentlyworking on to develop young talent.

• The Working Groups. I want to revive the former workinggroups by setting up new working groups on hot topicssuch as blockchain. This will allow us to really benefitfrom our mutual collaboration within ERCIM by coordi-

Jos baeten

Elected President

of ERCIM AISbL

The General Assembly of the ERCIMAISBL, held on 24 October in Lisbon,unanimously elected Jos Baeten, gen-eral director of Centrum Wiskunde &Informatica (CWI) in the Netherlandsas its new President for a period of twoyears as of January 2018. Jos Baetensucceeds Domenico Laforenza fromthe Institute for Informatics andTelematics (IIT) of the Italian NationalResearch Council (CNR) who servedas President of ERCIM since January2014.

Jos Baeten has a PhD in mathematicsfrom the University of Minnesota(1985). From 1991 to 2015, he wasprofessor of computer science at theTechnische Universiteit Eindhoven(TU/e). In addition, from 2010 to 2012he was professor of systems engi-neering at TU/e. As of October 2011,he is the general director of CentrumWiskunde & Informatica (CWI) inAmsterdam, the national researchinstitute for mathematics and computerscience in the Netherlands. Since

January 2015, he is part-time professorin theory of computing at the Instituteof Logic, Language and Computationof the University of Amsterdam. He iswell-known as a researcher in model-based engineering, in particular inprocess algebra.

Jos Baeten expressed his gratitude toDomenico for his dedication toERCIM during his four years of presi-dency. “The festive 25th anniversarycelebration of ERCIM at CNR, whereDomenico welcomed over a hundred

guests from academia, industry andpolitics, his efforts in approachingERCIM and Informatics Europe tocooperate, and his remarkablyinspiring speeches were highlightsduring Domenico’s presidency”, Jossays. Domenico began his activity inERCIM in 1993, contributing to thecreation of the ERCIM ParallelProcessing Network (PPN). He hasrepresented CNR on the Board ofDirectors since 2006 and will continuerepresenting CNR after his presidency.

Jos�Baeten�(left)�and�Domenico�Laforenza.


Joint ERCIM Actions

nating research in those new fields. Working groups havea limited life span, focusing on temporary, topical, short-term research that responds to new and current develop-ments. In terms of collaboration, it’s important thatresearchers, not just directors, will be able to find eachother. This is also the idea behind exchanging researchersin our Fellowships and PhD programme. Added value canbe found in new research areas.

ERCIM also successfully coordinates EU projects. This taskis carried out by the EEIG office, originally conceived as aservice for the members. I will strongly encourage the mem-bers to make use of it and cooperate through EU fundedresearch projects.

For our leading role in Europe, lobbying is also important. Iwould like to have more contact with Brussels. We must acttogether, ERCIM, Informatics Europe, and, for instance, theACM Europe Council and EIT Digital. The cooperatingorganisations must become members of advisory bodies inthe European Commission, such as the CONNECT AdvisoryForum for Research and Innovation in ICT in Horizon 2020.I commit myself to recruiting members and to reinforce con-tacts with the EC. Together, as an association of organisa-tions, we will endeavour to exert more influence on the poli-cies of the EU. For instance: to be heard at the creation of theEuropean Innovation Council (EIC) and to be involved in thepreparations for FP9, the programme that follows H2020. Weneed to be putting more effort into this area, and I intend tomake this my personal mission.

In addition to lobbying, collaboration is also important. Forinstance, I believe that the annual Informatics Europemeeting should continue to coincide with ERCIM meetings.This requires much consultation and coordination. The othersemi-annual meeting of ERCIM should also be forresearchers, and not just the directors.

One of ERCIM’s unique features is our combination of infor-matics and mathematics, and these areas are becomingincreasingly important for Europe. My personal opinion isthat Europe is not doing well: ICT companies are being soldto the US and China, which means that we no longer havecontrol over our data. I think this is an unhealthy trend; citi-zens must have control over their own data and at least knowwho has them. We do not necessarily have to own our databut we must be able to control who has or uses them.Together, we need to work to benefit all European citizens.Together, we can conduct research that will bring us theinnovations of the future – for the short and the long term. Itis vital for the future of Europe that we are actively involvedin helping to shape future innovations: that we do not lose theinnovation capacity, as we are already losing ground toemerging economies and other world powers. China and theUSA are out in front at the moment, and I think Europe musthave a leading role, certainly in the field of long-termresearch for future innovations.

With ERCIM, I hope that we can contribute to this goal.

Jos Baeten,

General Director of CWI and President of ERCIM AISBL

tim baarslag Winner

of the 2017 Cor baayen

Young Researcher Award

Tim Baarslag from CWI was selected as the winner of the

2017 ERCIM Cor Baayen Award, in a very tough

competition with 15 finalists. The award committee

recognises Tim’s skills and the results that he has

achieved. His enthusiasm for internationally oriented

research cooperation, his talent for recognising the

potential use of mathematical tools, and his cooperative

skills make him a young researcher of outstanding quality.

Tim has obtained some very impressive scientific results. HisPhD dissertation was awarded cum laude, which is conferredto less than 5% of doctoral candidates at the Delft Universityof Technology. His dissertation won the 2014 Victor LesserDistinguished Dissertation Runner-up Award in recognitionof an exceptional and highly impressive dissertation in thearea of autonomous agents and multiagent systems. He wasshortlisted for the 2014 Artificial Intelligence DissertationAward, which recognises the best doctoral dissertations inthe general area of artificial intelligence. His dissertation isalso published by the Springer Theses “the best of the best”series, which recognises outstanding PhD research byselecting the very best PhD theses from around the world fortheir scientific excellence and high impact on research.

Tim has made a real scientific impact with his work sinceembarking on his PhD in 2010. He has already publishedover 40 articles, in collaboration with 35 researchers from 17international institutes and five industrial partners. Hisresearch is published in the highest ranking publications inhis field, including the top journal Artificial Intelligence andtop conferences such as IJCAI, ECAI, AAAI and AAMAS.His research on how artificial intelligence can help negotiatebetter deals for humans was recently featured in Science. Hisachievements point to great promise for his future work andfor others who build on it.

Tim�Barslaag�(center)�receiving�the�award�from�the�ERCIM

president�Domenico�Laforenza�(right)�and�ERCIM�president-elect

Jos�Baeten.

Joint ERCIM Actions

Since 2016, Tim Baarslag has been working as a researcherat CWI, where he studies negotiation strategies for smartenergy cooperatives. He collaborates closely with key smartgrid stakeholders to develop his negotiation results into user-adaptable energy trading technology. His negotiation algo-rithms support energy trading in a sustainable energy com-munity ‘Schoonschip’ in Amsterdam as part of the ERA-NetGrid-Friends project. He has recently received a personal‘Veni’ grant for young, talented researchers from TheNetherlands Organisation for Scientific Research (NWO).

As the lead developer of the negotiation environment“Genius”, he contributed to an international platform forresearch on automated negotiation agents, which is down-loaded more than 100 times a week and is used by more than20 research institutes all over the world, including HarvardUniversity and Massachusetts Institute of Technology(MIT). Additionally, he is part of the organising team of theInternational Automated Negotiating Agent Competition(ANAC), which has had more than 100 international partici-pants and is held in conjunction with the leading artificialand multi-agent conferences in his field (AAMAS andIJCAI). Not only have his results and the competition pro-vided a state-of-the-art repository of automated negotiators,they also continue to steer the automated negotiationresearch agenda.

Tim’s work provides a unique blend of mathematicallyoptimal results and application-driven research with a strongpotential for knowledge utilisation. In collaboration withMIT CSAIL and the University of Southampton, he hasdeveloped new mathematical models that enable essentialnext steps for conducting privacy negotiations feasibly.Furthermore, together with multiple UK partners includingThe Open University, Tim helps to design Internet of Thingssolutions that enable negotiable data access for future datamarketplaces. His research has influenced the negotiationarchitecture that will be used to assess and mitigate privacytrade-offs in existing Internet of Things applications andplatforms.

Tim was the first to untangle the connection between negoti-ation performance and opponent learning. In acknowledge-ment of the originality of this work, he received the BestPaper Award (out of 101 papers) at the 2013IEEE/WIC/ACM International Conference on IntelligentAgent Technology. Moreover, he was the first to combine anumber of theoretical results in search theory, such asoptimal stopping and Pandora’s Problem, to formulate newand tested negotiation strategies. This resulted in the BestPaper Award (out of 122 papers) for his results on optimalnegotiation strategies during the 2015 IEEE/WIC/ACMInternational Conference on Intelligent Agent Technology.Moreover, a negotiating agent that successfully applied hisopen source dissertation framework won The 2013International Automated Negotiating Agents Competition(ANAC).

Tim is also actively engaged in the broader computing com-munity and highly involved with the most eminentresearchers within computer science and mathematics. In2016, he was selected as one of 200 highly talented, preemi-nent young researchers in mathematics and computer scien-tists from over 50 countries for the Heidelberg LaureateForum scholarship. In 2017, he was selected from over 300applicants as an outstanding early career individual in thefield of computing to take part in the ACM Future ofComputing Academy and attend the 2017 Turing Award cer-emony, in an effort to support and foster the next generationof computing professionals.

See also Tim’s article “Computers that Negotiate on OurBehalf” on page 34 in this issue.

Link:

ERCIM Cor Baayen Award: https://www.ercim.eu/human-capital/cor-baayen-award


ERCIM Cor baayen

Award 2017

Winner:Tim Baarslag (CWI), nominated byEric Pauwels (CWI)

Honorary mention:Fabrice Ben Hamouda-Guichoux(IBM Research), nominated by DavidPointcheval (Inria)

Finalists:• Markus Borg (RI.SE SICS),

nominated by Jakob Axelsson(RI.SE SICS)

• Frederik Diederichs (FraunhoferIAO), nominated by AnetteWeisbecker Fraunhofer (IAO)

• Hadi Fanaee Tork (University ofOslo), nominated by Alípio Jorge(INESC)

• Eemil Lagerspetz (University ofHelsinki), nominated by TuomoTuikka (VTT)

• Pierre Lairez (Inria), nominated byBertrand Braunschweig (Inria)

• Britta Meixner (CWI), nominatedby Pablo Cesar (CWI)

• Iason Oikonomidis (ICS-FORTH),nominated by Antonis Argyros(ICS-FORTH)

• Giulio Rossetti (University of Pisa),nominated by Fosca Giannotti(CNR)

• Ville Salo (University of Turku),nominated by Tuomo Tuikka (VTT)

• Anna Ståhl, RI.SE SICS, nominatedby Kristina Höök (RI.SE SICS)

• João Tiago Medeiros Paulo (INESCTEC and University of Minho),nominated byJosé Pereira (INESCTEC and University of Minho)

• Vassilis Vassiliades (Inria),nominated by Chris Christodoulou(University of Cyprus)

• Daniel Weber (Fraunhofer IGD),nominated by Arjan Kuijper(Fraunhofer IGD).


ERCIM “Alain bensoussan”

fellowship Programme

ERCIM offers fellowships for PhD holders from all overthe world. Topics cover most disciplines in ComputerScience, Information Technology, and AppliedMathematics. Fellowships are of 12 months duration,spent in one ERCIM member institute. Fellowships areproposed according to the needs of the member institutesand the available funding.

Application deadlines for the next rounds: 30 April

and 30 September 2018

More information: http://fellowship.ercim.eu/

HoRIZoN 2020

Project Management

A European project can be a richly rewarding tool forpushing your research or innovation activities to the state-of-the-art and beyond. Through ERCIM, our member instituteshave participated in more than 80 projects funded by theEuropean Commission in the ICT domain, by carrying outjoint research activities while the ERCIM Office success-fully manages the complexity of the project administration,finances and outreach.

The ERCIM Office has recognized expertise in a full rangeof services, including identification of funding opportunities,recruitment of project partners, proposal writing and projectnegotiation, contractual and consortium management, com-munications and systems support, organization of attractiveevents, from team meetings to large-scale workshops andconferences, support for the dissemination of results.

How does it work in practice?

Contact the ERCIM Office to present your project idea and apanel of experts will review your idea and provide recom-mendations. If the ERCIM Office expresses its interest toparticipate, it will assist the project consortium as describedabove, either as project coordinator or project partner.

Please contact:

Philippe Rohou, ERCIM Project Group [email protected]

ERCIM Established

Working Group

on blockchain technology

ERCIM established a new Working Group BlockchainTechnology to study the potential of this technology for arange of application fields in industry and public administra-tion. First chairperson of the Working Group is WolfgangPrinz from the Fraunhofer Institute for Applied InformationTechnology FIT.

Following up on the ERCIM workshop on BlockchainTechnology in May 2017 in Paris, ERCIM now establishedthe new Working Group Blockchain Technology. It hasbecome evident that blockchain technology is being investi-gated in various areas of computer science research. Thatincludes peer-to-peer networks, distributed systems, cryp-tography, algorithms for consensus building and validationas well as for modeling processes and business models. Inaddition to these basic technologies, the Working Group willalso study applications in relevant fields like Internet-of-Things, supply chains, energy and Smart Grid, the media, themedical field and the financial industry.

“We want to establish blockchain technology as a new fieldof computer science research in Europe. Our working groupwill strengthen the young community of Europeanblockchain researchers and foster interdisciplinary coopera-tion and exchange”, Wolfgang Prinz stated.

The Working Group will also establish a roadmap docu-menting ongoing research and open research questions in therapidly moving field of blockchain technology. Another ele-ment of its mission is to build a blockchain research commu-nity that will be able to successfully initiate and carry outcollaborative research projects on a European level.

For its first year the group is planning several meetings andeditorial projects in accordance with the objectives of thegroup. The next major event will be a workshop organized bymembers of the group, to be held May 8-9, 2018 inAmsterdam, in conjunction with the ERCIM spring meet-ings.

Researchers from ERCIM member institutions and otherorganizations are cordially invited to join the new ERCIMworking group Blockchain Technology.

Link:

ERCIM Blockchain WG: https://wiki.ercim.eu/wg/BlockchainTechnology/

Please contact:

For additional information please contact the workgroupchair Wolfgang Prinz Fraunhofer FIT, [email protected]

The development of this technology hasboth a hardware and a software compo-nent, both of which have been the sub-jects of intense research for the past fewdecades. On the hardware side, pastefforts went into controlling and storingsmall numbers (5-10) of qubits, the fun-damental building blocks for quantumcomputation. The current state of the artis that this can be done for arrays ofabout fifty qubits. Although a seeminglymodest scale, this is where thingsbecome very interesting, becausequantum systems of 50-100 qubitscannot be simulated on our current clas-sical computers and so hold the possi-bility of harnessing a computationalpower we have not seen before. A morelong-term goal (in the order of a fewdecades) is to scale-up further to mil-lions of qubits. This scale is what theNSA is worried about, because quantumcomputers of this size could break mostmodern-day cryptography.

Hardware, however, is not much goodwithout interesting software. Someimportant research areas on this side ofthe development include:• Quantum simulation, which includes

research on applications of medium-sized quantum computers. Potentialapplication areas include chemistryand materials science.

• Algorithms and complexity, whichexplores what larger-scale quantum

computers could do. Think forinstance about machine learning,search and optimisation and tacklingnumber-theoretic problems (relevantto cryptography).

• Cryptography, important becauselarger scale quantum computers willbreak our current cryptosystems, butalso hold the key for future cryp-tosystems.

• Quantum software framework,

including quantum programminglanguages, together with formal andlogical methods to ease the use ofquantum computers and understandthe fundamental structures of quan-tum information processing.

• Quantum information science in gen-

eral, which is to provide the mathe-matical theory behind quantuminformation processing and is impor-tant for the development of error cor-rection schemes, understandingcounterintuitive quantum phenome-na and the physical theory behind thehardware.

The articles in this special theme offer amore detailed look into the complexi-ties of some of the above-mentionedaspects of the emerging paradigm shiftcurrently happening in computation.Below you will find lightning-fastintroductions to the basic conceptsappearing in the articles.

Basic features of quantumcomputationQubits and superpositions

The basic building block for classicalcomputation is the bit, a physicalsystem that can be toggled between oneof two possible states, 0 and 1. Forquantum computers, this building blockis fundamentally different. It is thequbit, short for quantum bit. A qubit is aphysical system that can be in a sort ofintermediate state given by one of theinfinitely many possible superpositionsof two basis states. If the two basisstates are represented by two-dimen-sional row vectors (1,0) and (0,1), thena superposition is a complex Euclideanunit vector of the form (x,y) where x andy are complex numbers, referred to asprobability amplitudes, whose absolutevalues squared sum to 1. A measure-ment of the qubit results in finding it inthe first or second state with probability|x|2 or |y|2, respectively. Toggling ofquantum states can be done via linear,length-preserving operations, in otherwords, unitary transformations. A meas-urement performed after a unitary oper-ation is also referred to as doing a meas-urement in a different basis. Whereasthe state of n classical bits can be repre-sented by an n-bit string, the state of ann-qubit system is in general given by asuperposition of 2n basis states: a com-plex 2n-dimensional Euclidean unitvector. The observation that n-qubit


Special theme: Quantum Computing

Introduction to the Special Theme

Quantum Computation and Information

by Jop Briët (CWI) and Simon Perdrix (CNRS, LORIA)

For more than a century now, we’ve understood that we live in a quantum world. Even though quantum

mechanics cannot be ignored during the development of atomic scale components of everyday

computers, the computations they perform are governed, like the Turing machine, by the laws of classical

Newtonian mechanics. But the most striking and exotic features of quantum mechanics, superposition

and entanglement, currently play no part in every-day information processing. This is about to change –

and in some specialised applications, already has. In academia, the field of quantum computation has

been growing explosively since its inception in the 1980s and the importance of these devices is widely

recognised by industry and governments. Big players in the tech industry like IBM and Google frequently

announce that they have built yet a larger rudimentary quantum computation device and in 2016 the

European Commission launched a one-billion Euro Flagship Initiative on Quantum Technologies.

states require an exponential number ofparameters to be described is whatmakes simulating quantum-mechanicalsystems hard to do on classical com-puters. It is also one of the key featuresof quantum mechanics that givesquantum computers their power.

Entanglement

Arguably the most striking features ofquantum mechanics is entanglement,which manifests itself when two ormore quantum systems are measuredlocally in one of two or more possiblebases. As an argument against quantum,it was observed by Einstein, Podolskyand Rosen that compound systemsallow superpositions that can result inmeasurement statistics that defy clas-sical explanation. Celebrated work ofBell later showed that one can actuallytest this feature experimentally usingwhat is nowadays referred as a Bell test.The most basic example of such a testcan be cast as a game involving twoplayers, Alice and Bob, and a referee.The referee picks two bits at randomand sends these to Alice and Bob,respectively. Without communicating,and thus without knowing the other’sbit value, the players each return abinary answer to the referee. They winthe game if the sum of the answersmodulo 2 equals the product of thequestions. A simple calculation showsthat if they players are constrained bythe laws of classical physics, then theycan win this game with probability atmost 3/4. However, by performing localmeasurements on a two-qubit “EPRpair”, the players can produce a proba-bility distribution over their answerpairs with which they win with proba-bility roughly 0,85!

Quantum algorithms

Quantum algorithms consist of a care-fully chosen sequence of unitary trans-formations that are applied one by oneto a register of qubits, followed in theend by a measurement to determine anoutput. A crucial property of unitarytransformations is that they can causecancellations among the probability

amplitudes describing the overall stateof the register. These cancellations canin turn cause large superpositions toquickly converge into a state that, whenmeasured, with near-certainty givesonly a single possible outcome. Twoearly, but still important, examplesdemonstrating the power of this phe-nomenon are Shor’s factoring algorithmand Grover’s search algorithm. Shor’salgorithm factors a given number intoits prime-number components exponen-tially faster than the best-known clas-sical algorithm to date, with dramaticconsequences for important crypto-graphic schemes (see below). Grover’salgorithm finds the location of a 1 withgood probability in a given n-bitsequence, provided there is one, using anumber of basic computational stepsgiven by roughly the square root of n.The best-possible classical algorithmneeds roughly n steps in the worst case,however.

Quantum cryptography

Shor’s algorithm can break the mostimportant cryptographic schemes wehave today, which are based on theassumption that factoring is hard. Afully operational large-scale quantumcomputer shatters this assumption. Thismeans that alternative cryptography isneeded immediately, as some of today’sinformation may need to be kept secreteven after quantum computers are built.Multiple lines of research on post-quantum cryptography address thisissue. On the one hand, one can try tolook for problems that might even behard for quantum computers, an impor-tant motivation for lattice-based cryp-tography. On the other hand, quantummechanics itself offers alternatives too,as was pointed out already long beforeShor’s discovery. Wiesner circa 1970introduced a quantum money scheme, aproposal where bank notes are unfalsifi-able thanks to the presence of qubits onit. This was the first attempt to usequantum mechanics in a cryptographicprotocol. More than a decade later, in1984, Bennett and Brassard introducedthe revolutionary quantum key distribu-

tion protocol, an unconditionally securecommunication protocol relying on thelaws of quantum mechanics. In this pro-tocol, Alice wants to share a randomkey with Bob, to do so she sendsrandom bits encoded into randomlychosen basis among two complemen-tary basis. Complementary basis aresubject to the uncertainty principle: if abit of information is encoded into one ofthe basis, measuring according to theother one produces an random bit,uncorrelated with the encoded informa-tion. Roughly speaking a spy who wantsto observe the sent qubits will not onlyget no information if he does not guessthe appropriate basis, but will bedetected by Alice and Bob with highprobability. The protocol is relativelysimple to implement, and has been com-mercialised. Notice however, the proofthat the protocol is actually uncondi-tionally secure took 15 years!

Scalability and error correction

One of the major challenges in the questfor a quantum computer is its scala-bility. Prototypes of quantum computersalready exist but their memory, subjectto decoherence, is limit to a few dozenqubits. The scalability of quantum com-puters is not only a technological chal-lenge, error correcting code are crucial:a large scale robust quantum computerrequires that the quality of the quantumdevice should meet the maximalamount of errors quantum codes cancorrect.

Please contact:

Jop BriëtCWI, The [email protected]

Simon PerdixLORIA, CNRS, Inria Mocqua,Université de Lorraine, [email protected]


It is already possible to buy quantumhardware for certain cryptographictasks, such as random number genera-tion and key distribution. In crypto-graphic scenarios, being able to test thatyour devices behave as advertised isclearly of paramount importance.

The first commercial quantum com-puters are likely to run as shared servers,where clients can pay to have a quantumcomputation run. This is another sce-nario where we would like the client,who does not have a quantum computer,to have some guarantee that the correctcomputation has been performed.

These scenarios are captured by a verifi-

cation protocol, in which a verifier, whowould like to run some quantum compu-tation, interacts with one or moreprovers who have a quantum computer.By the end of the protocol, the verifiereither “accepts” or “rejects” the outputof the computation. The verifier mightrepresent an experimenter testing aquantum system, and the provers a per-sonification of nature, or more precisely,

the physical systems being tested. If theprovers are “honest” and follow the pro-tocol – that is, they behave as predicted– the verifier should learn the result ofthe quantum computation. However, ifthe provers are deviating from the pro-tocol, the verifier should detect this and“reject”.

Verifying quantum computations isrelated to the more fundamental task ofexperimentally verifying quantummechanics. A quantum system has aninternal state that an observer cannotperceive directly. To learn about thisstate, a quantum measurement can beperformed, giving some incompleteinformation. If an experimentalisthypothesises that a particular quantumsystem has a particular state, this can beverified by performing measurements,but this presupposes some trustedquantum measurement device. If onewants to verify the theory of quantummechanics, one cannot circularlyassume that the measurement devicebehaves as predicted by the theoryquantum mechanics.

However, there are means of verifyingcertain aspects of quantum mechanicsthat do not require the experimenter tohave a trusted quantum device, calledBell tests. In a Bell test, two provers playa game with a verifier. The verifier askseach prover a question, and they musteach return an answer, without commu-nicating during the game. What makes aBell test special is that they can win thegame with higher probability if theyshare a quantum resource called entan-glement than if they are classical. Thus,such a game offers a way to experimen-tally test quantum mechanics.

Analogous to the situation in testingquantum mechanics, testing a quantumcomputer can be accomplished in eitherof two regimes. In the quantum-verifier

regime, the verifier must have a simpletrusted quantum device, like a measure-ment device. In this regime, several effi-cient verification protocols are known.However, the requirement that the veri-fier have a trusted quantum device is abig drawback.

In the two-prover regime, analogous toa Bell test, a classical verifier interactswith two provers who do not communi-cate. The first protocol in this regimewas a significant breakthrough, pro-viding, for the first time, a method for aclassical verifier to verify any quantum



How Classical beings

Can test Quantum devices

by Stacey Jeffery (CWI)

As the race to build the first quantum computer heats up, we can soon expect some lab to claim to

have a quantum computer. How will they prove that what they have built is truly a quantum

computer?

Verifier Prover

Prover V Prover P

Verifier

Figure�1:�Quantum-verifier�Regime.�A�verifier�with�a�simple�quantum

device�interacts�with�a�prover�with�a�full�quantum�computer.

Figure�2:�Two-prover�Regime.�A�classical�verifier�interacts�with�two

quantum�provers.

Figure�3:�Our�new�two-prover�protocol.�The�provers�play�the�role�of�the

verifier�and�prover�from�Broadbent’s�Protocol,�which�is�a�quantum-

verifier�protocol.

computation [3]. However, the protocolhad the major disadvantage of requiringresources that scale like g8192 to verifi-ably implement a quantum circuit ofsize g. For comparison, there are an esti-mated 2286 elementary particles in theuniverse. Subsequent improvementsdecreased this overhead to g2048, butthis is still thoroughly impractical, evenfor a quantum circuit of size g = 2.

In collaboration with researchers fromUniversité Paris Diderot and Caltech,we presented the first efficient protocolin the two-prover regime [2]. This pro-tocol requires resources that scale likeO(g log g) to verify a quantum circuit ofsize g.

We begin with a quantum-verifier pro-tocol due to Broadbent [1]. The proversin our protocol are asked to simulateBroadbent's verifier and prover, respec-

tively. We call our provers Prover V, forverifier, and Prover P, for prover.

By the properties of Broadbent’sProtocol, we can use Prover V to test

Prover P, to make sure he is followingthe protocol. Then it only remains toensure that Prover V is, in fact, fol-lowing the protocol. We develop newBell tests that are used to make Prover Ptest Prover V. While the classical verifiermay not be powerful enough to keep aquantum prover in check, she can usethe two provers to control one another.

In the future, we hope to see experi-mental realisations of our protocol,which is possible for the first time, dueto its efficiency. Its near-optimal effi-ciency means that verifying a particularquantum computation does not requiremuch more resources than simplyimplementing that computation.

Links:

CWI’s algorithms and complexitygroup: https://kwz.me/hBD QuSoft: http://www.qusoft.org/

References:

[1] A. Broadbent: “How to Verify aQuantum Computation” Arxiv preprintarXiv:1509.09180, 2015.[2] A. Coladangelo, A. Grilo, S.Jeffery, T. Vidick: “Verifier-on-a-Leash: new schemes for verifiabledelegated quantum computation, withquasilinear resources” Arxiv preprintarXiv:1708.07359, 2017.[3] B. W. Reichardt, F. Unger and U.Vazirani: “Classical command ofquantum systems” Nature 496, 456-460, 2013.

Please contact:

Stacey Jeffery, CWI and QuSoft,[email protected]


Quantum information theory has radi-cally altered our perspective aboutquantum mechanics. Initially, researchinto quantum mechanics was devoted toexplaining phenomena as they areobserved in nature. But the focus thenchanged to designing and creatingquantum systems for computation,information processing, communica-tion, and cryptography among manyother tasks. In particular, what becameclear was that quantum interference -“the heart of quantum mechanics”, asRichard Feynman described it - can beharnessed for quantum computation.Algorithms running on a hypotheticalquantum computer would be able tosolve problems by creating an interfer-ence pattern of different computationalbranches. This can lead to an exponen-tial saving in the amount of resourcesused by a quantum algorithm, whencompared to the best known classicalalgorithms. The most famous exampleof this is Shor's algorithm for factoring

numbers which is exponentially fasterthan the best known classical factoringalgorithms.

But having a device which can solveproblems exponentially faster than clas-sical computers raises an interestingquestion: can a classical computer effi-ciently verify the results produced bythis device? At first, one might betempted to dismiss this question and saythat as long as each component of aquantum computer has been tested andworks correctly, there is no need toworry about the validity of the device'sresults. However, the point of verifica-tion is much more profound. Quantumcomputers would provide one of themost stringent tests of the laws ofquantum mechanics. While numerousexperiments involving quantum sys-tems have already been performed to aremarkable precision, they all utilizedrelatively few degrees of freedom. Butwhen many degrees of freedom are

involved, and because predicting theoutcome of the experiment requiresexponential resources, it quicklybecomes infeasible to calculate the pos-sible results of the experiment withoutresorting to lax approximations.Verification of quantum computationwould therefore allow for a new test ofquantum mechanics, a test in the regimeof high complexity.

There is another important reason forverifying quantum computations,having to do with cryptography. Thefirst quantum computers are likely to beservers, to which clients can connectthrough the Internet. We can already seean instance of this with the recent 5-qubit and 16-qubit devices that IBM hasmade available to the general public[L1]. When larger devices becomeavailable, users will wish to delegatecomplex computations to them.However, in such a distributed environ-ment, malicious agents might perform

Keeping Quantum Computers Honest

(or verification of Quantum Computing)

by Alexandru Gheorghiu (University of Edinburgh) and Elham Kashefi (University of Edinburgh, CNRS)

Quantum computers promise to efficiently solve not only problems believed to be intractable for

classical computers, but also problems for which verifying the solution is also intractable. How then,

can one check whether quantum computers are indeed producing correct results? We propose a

protocol to answer this question.

man-in-the-middle attacks or compro-mise the remote server. The clientswould then need a means to check thevalidity of the server's responses. Infact, in this setting, users might alsowish to keep their data hidden evenfrom the quantum computer itself, as itmight involve sensitive or classifiedinformation.

So can one verify quantum computa-tions while also maintaining the secrecyof the client's input? The answer is yes.In fact, the client's ability to keep theinput hidden is what makes verificationpossible. This was shown by Fitzsimonsand Kashefi when they proposed a veri-fication protocol based on a crypto-graphic primitive known as UniversalBlind Quantum Computation (UBQC)[1,2]. In UBQC, a client that can pre-pare single qubits has the ability to dele-gate quantum computations to a server,in such a way that the server is obliviousto the computation being performed. Todo verification, the client can thenexploit this property by embedding testsin the computation, referred to as traps,which will fail if the server doesn't per-form the correct computation. Ofcourse, the problem with this approachis that the client needs to trust that thequbit preparation device works cor-rectly and produces the specified states.But if, prior to the start of the protocol, amalicious agent corrupts the prepara-tion device, the client could later betricked into accepting incorrect results.

To address this issue, we, together withDr. Petros Wallden, at the University ofEdinburgh, proposed a verification pro-tocol which is device-independent [3].In other words, the client need not trustany of the quantum devices in the pro-tocol. This is achieved by using a pow-erful result of Reichardt, Unger and

Vazirani, known as rigidity of non-localcorrelations [4]. Non-local correlationsare correlations between responses ofnon-communicating parties that cannotbe reproduced classically, unless theparties are allowed to communicate.Such correlations can be produced,quantum mechanically, through a suit-able strategy for measuring certainentangled states. The rigidity result isessentially a converse to this. It statesthat certain non-local correlations canonly be produced by a particular, uniquestrategy. Observing such correlationsbetween non-communicating devicesthen implies that the devices arebehaving according to this fixedstrategy. What is remarkable about thisresult is that it only requires examiningthe outputs of the devices, withoutassuming anything about their innerworkings.The protocol then works as follows: theclient has an untrusted device for meas-uring single qubits and is also commu-nicating classically with the quantumserver. By examining the outputs of thetwo devices, it follows from the rigidityresult that the client can check whetherthe two devices are sharing entangle-ment and performing measurements asinstructed. If so, the client leverages thisand uses the entanglement to remotelyprepare single qubit states on theserver's side. Finally, the client uses thetrap-based scheme of Fitzsimons andKashefi to delegate and verify an arbi-trary quantum computation to theserver.

Verification is an important milestoneon the road to scalable quantum com-puting technology. As we have seen,verification protocols exist even for themost paranoid users. But even so, ques-tions still remain regarding their opti-mality, their ability to tolerate noise and

imperfections, as well as other issues.Addressing all these questions is a keychallenge for both theorists and experi-mentalists and their resolution willshape the landscape of the emergingQuantum Internet.

Link:

[L1] https://kwz.me/hBv

References:

[1] A. Broadbent, J.F. Fitzsimons, E.Kashefi: “Universal blind quantumcomputation”, in Proc. of FOCS‘09, IEEE Computer Society(2009) 517 – 526.

[2] J.F. Fitzsimons, E. Kashefi:“Unconditionally verifiable blindquantum computation”, Phys. Rev.A 96 (2017) 012303.

[3] A. Gheorghiu, E. Kashefi, P.Wallden: “Robustness and deviceindependence of verifiable blindquantum computing”, New Journalof Physics 17(8) (2015) 083040.

[4] B.W. Reichardt, F. Unger, U.Vazirani: Classical command ofquantum systems. Nature496(7446) (2013) 456.

Please contact:

Elham Kashefi, University ofEdinburgh, UK and CNRS, [email protected]

Alexandru GheorghiuUniversity of Edinburgh, [email protected]



Figure�1:�Device-independent�verification�protocol.�The

client,�or�verifier,�will�instruct�both�the�measurement�device

and�the�server�to�measure�entangled�qubits.�The�statistics�of

these�measurements�are�then�checked�by�the�verifier.�All

communication�with�the�quantum�devices�is�classical.


The prospect of harnessing preciselyengineered quantum systems to runalgorithms which vastly outperformtheir best classical counterparts has gen-erated a large amount of interest andinvestment in the field of quantum com-puting. However, a universal and fault-tolerant quantum computer is unlikely tobe built soon. Considering this poten-tially long wait, recently several newproblems have been proposed with thepurpose of demonstrating “quantumcomputational supremacy”, which refersto the point at which a quantum machineperforms a computational task that isbeyond the capability of any classicalcomputer [1], with specialised near-termquantum devices.

One such proposal is the boson samplingproblem introduced by Aaronson andArkhipov [2]. The problem consists ofsampling from the output distribution ofdetection events generated when manysingle photons are concurrently injectedin to a randomly chosen network oflinear optical components. A samplevalue is generated by recording wherephotons were detected at the end of thelinear optical network. The probabilityfor each possible output in the experi-ment is related to a matrix functionknown as the permanent, which is ingeneral especially hard to compute. Bymaking the plausible conjecture that thepermanent is hard to approximate forGaussian random matrices, as well asanother reasonable conjecture about thedistribution of these permanents,Aaronson and Arkhipov were able toshow that an efficient classical algo-rithm for even approximate Boson sam-pling would lead to the collapse of thepolynomial hierarchy of complexityclasses to its third level – somethingconsidered extremely unlikely in com-putational complexity theory. The resultapplying to approximate boson sam-pling is key, as it allows for the possi-bility of solving the boson samplingproblem with realistic quantum experi-ments, without the need for quantum

error correction. Physically, the com-plexity of the problem stems from thecomplex quantum interference of indis-tinguishable photons; if we make thephotons distinguishable (say, by usingphotons of different colours) then theproblem is no longer hard.

Our best new classical boson samplingalgorithm [3] is based on MetropolisedIndependence Sampling (MIS), aMarkov chain Monte Carlo proceduresimilar to the famous Metropolis-Hastings algorithm. We start a list byproposing a sample value from someeasy-to-sample distribution, and con-tinue by iteratively proposing newvalues and either adding this new valueto the list or repeating our previousvalue. Which of these possibilities actu-ally occurs is determined by a carefullycrafted acceptance rule which dependson the probabilities of the proposed andcurrent value occurring in both thetarget distribution and the proposal dis-tribution, and guarantees that the listtends towards a genuine boson sam-pling sample.

We identified the corresponding distri-bution for distinguishable particles atthe output of the linear optical networkas a proposal distribution which pro-vides rapid convergence to the bosonsampling distribution. Although theprobabilities in this distribution are alsogiven by matrix permanents (albeit dif-ferent matrices), the distribution can beefficiently sampled. Using this, wefound strong numerical evidence thatcomputing just 200 matrix permanentsis enough to generate a boson samplingvalue via MIS for a problem size of upto 30 photons, roughly amounting to aspeed-up of 50 orders of magnitudeover the best previously suggested clas-sical algorithm at this problem size.By assuming that 200 matrix permanentcomputations suffice to produce a goodsample for larger photon numbers, wewere able to predict the amount of timeit would take to solve boson sampling

for up to 100 photons if we ran MIS on apowerful supercomputer. Alongsidethis, we computed the expected time toexperimentally perform boson samplingas a function of the probability of asingle photon surviving the experiment(i.e., not getting lost from source todetector). We found that it wouldrequire more than 50 photons before theclassical runtime would exceed a week.To achieve this experimentally wouldnot only require the ability to produce astate of 50 indistinguishable photons ata high rate (the current record-holdingboson sampling experiment is with fivephotons), but also that all photons sur-vive the experiment with a probabilitygreater than 50%. Considering that eachphoton would require to be injected into the circuit, pass through over 1000beamsplitters and be coupled out of thecircuit to single photon detectors, thiswould amount to a significant engi-neering breakthrough which is unlikelyto be realised in the near future.Our results not only provide a bench-mark for quantum supremacy by bosonsampling, but also highlight how signif-icant the experimental improvementmust be in order to achieve this.

Link:

https://kwz.me/hB1

References:

[1] A. Harrow, A Montanaro: “Quantumcomputational supremacy”, Nature549 (2017), 203–209.

[2] S. Aaronson, A. Arkhipov: “The com-putational complexity of linear optics”,Theory Comput. 9, 143–252, 2013.

[3] A. Neville, C. Sparrow, et al.:“Classical boson sampling algo-rithms with superior performance tonear-term experiments”, NaturePhysics 13, 1153–1157 (2017).

Please contact:

Alex Neville, Chris SparrowUniversity of Bristol, [email protected],[email protected]

Experimental Requirements for Quantum

Computational Supremacy by boson Sampling

by Alex Neville and Chris Sparrow (University of Bristol)

Boson sampling has emerged as a leading candidate for demonstrating “quantum computational supremacy”.

We have devised improved classic al algorithms to solve the problem, and shown that photon loss is likely to

prevent a near-term demonstration of quantum computational supremacy by boson sampling.

Information theory is the area of com-puter science that develops and studiesabstract mathematical measures of“information” – or, from a more pes-simistic perspective, measures of“uncertainty”, which capture the lack ofinformation. It is clear that when youtoss a coin there will be uncertainty inthe outcome: it can be either “head” or“tail”, and you have no clue what it willbe. Similarly, there is uncertainty in theface that will show up when you throw adice. It is even intuitively clear thatthere is more uncertainty in the latterthan in the coin toss. However, suchcomparisons become less clear for morecomplicated cases. If we want to com-pare, say, tossing three coins on the onehand with throwing two dice and takingthe sum of the two faces on the otherhand, it is not immediately clear inwhich of the two there is more uncer-tainty: there are fewer possible out-comes in the former, namely eight, but,on the other hand, the eleven possibleoutcomes in the latter are biased.

Information theory offers quantitativemeasures that express precisely howmuch uncertainty there is. Similarly, italso offers measures of conditionaluncertainty given that one holds some“side information”. For instance, howmuch uncertainty is there in the faces ofthe two dice given that I know the sumof the two? How much uncertainty isthere in a message that was communi-cated over a noisy channel given that Ihold the received noisy version? Howmuch uncertainty is there in a digitalphoto given that I hold a compressedversion? How much uncertainty does aneavesdropper have on secret data giventhat he got to see an encryption?Information theory allows us to answersuch questions in a precise manner andto make rigorous predictions about thebehaviour of information in all kinds ofinformation processing tasks. As such,information theory had – and still has – amajor impact on the development of

today’s information and communicationinfrastructure.

What makes information theory verypowerful is its independence of howinformation is physically represented:whether the information is representedby coins that show “head” or “tail”, orby the tiny indentations on a DVD, orwhether the information is stored on aflash drive or communicated overWiFi, the predictions of informationtheory hold universally – well, until wehit the realm of quantum mechanics. If,say, we encode information into thepolarisation of photos, then informa-tion starts to behave very differently.Therefore, a quantum version of infor-mation theory is necessary in order torigorously study the behaviour of infor-mation in the quantum realm, and, forinstance, to be able to quantify theamount of information an eavesdroppermay have on secret data when the datais protected by means of quantum cryp-tography, or to quantify the amount oferror correction needed in order tocounter the loss of information inquantum computation caused by deco-herence.

From a mathematical perspective, giventhat quantum mechanics is described bynon-commuting mathematical objects,quantum information theory can beunderstood as a non-commutative

extension of its classical commutativecounterpart. This insight can serve as aguideline for coming up with quantumversions of classical information meas-ures, but it also shows a typical predica-ment: a commutative expression can begeneralised in various ways into a non-commutative one. For instance, anexpression like A⁵B⁴ can be generalisedto A⁵B⁴ or to B⁴A⁵ for non-commutingA and B, or to ABABABABA, or toB²A⁵B², etc. One of the challengingquestions is to understand which of thepossible generalisations of classicalinformation measures are suitablemeasures of quantum information andhave operational significance.

Building upon new insights and newresults [1] that we discovered on theclassical notion of Rényi entropy, and incollaboration with several partners, wesucceeded in lifting the entire family ofRényi entropies to the quantum setting[2,3]. The Rényi entropies form a con-tinuous spectrum of information meas-ures and cover many important specialcases; as such, our extension to thequantum setting offers a whole range ofnew quantum information measures.We showed that our newly proposeddefinition satisfies various mathemat-ical properties that one would expectfrom a good notion of information andwhich make it convenient to work withthe definition. It is due to these that our



from Classical to Quantum Information –

or: When You Have Less than No Uncertainty

by Serge Fehr (CWI)

Over the last few years, significant progress has been made in understanding the peculiar behaviour

of quantum information. An important step in this direction was taken with the discovery of the

quantum Rényi entropy. This understanding will be vital in a possible future quantum information

society, where quantum techniques are used to store, communicate, process and protect information.

quantum Rényi entropies have quicklyturned into an indispensable tool forstudying the behaviour of quantuminformation in various contexts.

One very odd aspect of quantum infor-mation is that uncertainty may becomenegative: it may be that you have lessthan no uncertainty in your target ofinterest. This peculiarity is an artifactof entanglement, which is one of themost bizarre features of quantummechanics. Entanglement is a form ofcorrelation between quantum informa-tion that has no classical counterpart. Asensible explanation of negative uncer-tainty can be given as follows. ByHeisenberg’s uncertainty principle,even when given full information in theform of a perfect description of thequantum state of interest, there is stilluncertainty in how the state behaves

under different measurements.However, if you are given anotherquantum state that is entangled with thestate of interest, then you can actuallypredict the behaviour of the state ofinterest under any measurement bymeans of performing the same meas-urement on your state. Indeed, by whatEinstein referred to as “spooky actionat a distance”, the measurement onyour entangled state will instanta-neously affect the other state as to pro-duce the same measurement outcome.

The above aspect nicely illustrates thatquantum information theory is muchmore than a means for understandingthe behaviour of information withinpossible future quantum communica-tion and computation devices: it shedslight on the very foundations ofquantum mechanics itself.

References:

[1] S. Fehr, S. Berens: “On theConditional Rényi Entropy”, inIEEE Trans. Inf. Theory60(11):6801 (2014).

[2] M. Müller-Lennert, F. Dupuis, O.Szehr, S. Fehr, M. Tomamichel:“On Quantum Rényi Entropies: ANew Generalization and SomeProperties”, in J. Math. Phys54:122203 (2013).

[3] M. Wilde, A. Winter, D. Yang:“Strong Converse for the ClassicalCapacity of Entanglement-Breaking and Hadamard Channelsvia a Sandwiched Rényi RelativeEntropy”, in Commun. Math. Phys331:593 (2014).

Please contact:

Serge Fehr, CWI, The Netherlands +31 20 592 42 57, [email protected]


Most current day cryptosystems arebased on two computationally hardproblems: factoring integers and com-puting the discrete logarithms in finitegroups. Advances in solving these prob-lems (classically) and increased clas-sical computational power form a threatthat is relatively easy to diminish by‘simply’ increasing the key sizes. Amore serious threat was revealed in 1994already, when Shor [1] had proven thatquantum algorithms are able to effi-ciently solve both these problems. Whilethe requirement for a truly universalquantum computer to execute theseattacks has long kept us secure, recentadvances in quantum computing onceagain highlight our dependency on thesecomputational problems. In short, thearrival of a quantum computer will leavecommonly used cryptosystems such asRSA, DH and ECDH insecure.

To ensure secure communication in thepresence of a universal quantum com-puter, for many years cryptographershave been working on constructingcryptographic schemes based on othercomputational problems since manyyears. Their efforts have led to the fol-lowing six possible building blocks forcryptographic schemes: hash trees,error-correcting codes, lattices, multi-variate equations, supersingular ellipticcurve isogenies, and even quantumphysics. These cryptographic buildingblocks fall within the realm of so-called“post-quantum cryptography”, thestudy of cryptographic algorithms thatare secure against attacks by a quantumcomputer.

PROMETHEUS is a new, four-yearEuropean H2020 project (starting inJanuary 2018) aiming to provide a securedesign and implementation of new and

quantum-safe cryptographic systems.The twelve project partners (ENS Lyon,ORANGE SA, CWI Amsterdam, IBMResearch, RHU London, RU Bochem,Scytl Barcelona, Thales, TNO, UPCBarcelona, University Rennes, WISRehovot) will focus their efforts on lat-tice-based cryptography. A fundamentalproperty of lattice-based cryptographicschemes is that their security can bereduced to well-studied computationalproblems, which is not necessarily thecase for other post-quantum mecha-nisms.

An n-dimensional lattice L is a set ofinteger linear combinations of n inde-pendent vectors. The grid displayed inFigure 1 represents a lattice L. Anexample of a computationally hard lat-tice problem is the closest vectorproblem (CVP): given a lattice L and arandom point y (not necessarily an ele-

Preparing ourselves for the threats

of the Post-Quantum Era

by Thijs Veugen (TNO and CWI), Thomas Attema (TNO), Maran van Heesch (TNO), and Léo Ducas (CWI)

In the post-quantum era, most of the currently used cryptography is no longer secure due to

quantum attacks. Cryptographers are working on several new branches of cryptography that are

expected to remain secure in the presence of a universal quantum computer. Lattice-based

cryptography is currently the most promising of these branches. The new European PROMETHEUS

project will develop the most secure design and implementations of lattice-based cryptographic

systems. Exploitation of the project results will be stimulated by demonstrating and validating the

techniques in industry-relevant environments.

ment of the lattice), find the lattice ele-ment x closest to y. For higher dimen-sional lattices this problem soonbecomes very hard to solve. In fact, no(quantum) algorithms have been foundyet solving this problem efficiently.

Research in using this and other hardlattice problems for cryptographic pur-poses began with the publication of thepublic key encryption scheme by Ajtaiand Dwork in 1997 [2]. Improvementsto this scheme and new schemes fol-lowed, trying to make use of additionalstructures in specific types of lattices.The first lattice-based cryptographicschemes for public-key encryption, sig-natures and key-exchange have beenproposed, and are considered for stan-dardization [L1]. A recently proposedkey-exchange protocol [3] (partlydeveloped at CWI) has successfullybeen implemented and tested by Googlein the Chrome web browser [L2], andwas awarded the Facebook InternetDefense prize [L3].

One particular challenge for the transferfrom theory to practice lies in the choiceof parameters for those schemes: whilelattice problems become ‘hard’ withlarger dimensions, predicting preciselyhow hard they are (e.g., it will take 100years to solve with a cluster of 10,000GPUs) remains difficult and uncertain.The issue becomes even more delicatewhen considering quantum algorithms,especially in the light of recent quantumalgorithms specialised for “ideal lat-tices” [4]. This is the main issue forwhich CWI’s crypto group will provideits expertise.

The PROMETHEUS project recognisesthat modern day cryptography entails

much more than protecting privateinformation over insecure communica-tion channels. With its digital signa-tures, commitment schemes, and homo-morphic properties, lattice-based cryp-tography offers a wide variety of appli-cations. To enhance the applicabilityand the adaptation of these techniques,four use-cases will be studied. Bymeans of these use-casesPROMETHEUS aims to cover theentire range from theory to application.

The first use case will focus on con-structing an anonymous credentialsystem, which allows a user to prove toa service provider that he owns a certainattribute (e.g., driving licence), whileminimising the information given tothird parties, then protecting the user’sprivacy. In the second use case tech-nology will be developed that allowsusers to make secure, privacy-friendlycontactless transactions, for a long-termuse. In the third use case, a long-termsecure e-voting system will be devel-oped. Lastly, in the fourth use case,PROMETHEUS will develop quantum-safe homomorphic encryption tech-niques, and use them to develop asecure cyber threat-intelligence sharingmechanism.

All four use-cases aim for long-termsecurity in the quantum era, requiringthe cryptographic building blocks to bequantum safe. Many of these techniquesalready exist in a quantum vulnerablesetting. Introducing long-term securityby mitigating quantum threats is thecore of our innovation. By covering theentire range from theory to applicationand building demonstrators, theexploitation of the project results willbe stimulated.

Links:

[L1] https://kwz.me/hB2 [L2] https://kwz.me/hB3 [L3] https://kwz.me/hB5

References:

[1] P.W. Shor: “Algorithms forquantum computation: Discretelogarithms and factoring”, in Proc.of SFCS ‘94, 1994.

[2] M. Ajtai, C. Dwork: “A public-keycryptosystem with worst-case/average-case equivalence”,Proc. of the 29th annual ACMsymposium on Theory ofComputing. ACM, 1997.

[3] E. Alkim, L. Ducas, T.Pöppelmann, P. Schwabe: “Post-quantum Key Exchange-A NewHope”, USENIX SecuritySymposium, 2016.

[4] R. Cramer, L. Ducas, B.Wesolowski: ‘’Short StickelbergerClass Relations and application toIdeal-SVP’’, IACR, Eurocrypt2017.

Please contact:

Thijs Veugen, TNO, The Netherlands+31888667314, [email protected]



Figure�1:�Illustration�of

a�two-dimensional�lattice.�


Quantum Cryptography beyond Key distribution

by Georgios M. Nikolopoulos (IESL-FORTH, Greece)

Quantum cryptography is the science of exploiting fundamental effects and principles of quantum

physics, in the development of cryptographic protocols that are secure against the most malicious

adversaries allowed by the laws of physics, the “quantum adversaries”. So far, quantum cryptography

has been mainly identified with the development of protocols for the distribution of a secret truly

random key between two legitimate users, known as quantum key-distribution (QKD) protocols.

Beyond QKD, quantum cryptography remains a largely unexplored area. One of the main ongoing

projects at the Quantum Optics and Technology group of IESL-FORTH [L1], is the design and

development of cryptographic solutions, which rely on fundamental quantum-optical systems and

processes, and offer security against quantum adversaries.

Figure�1:�Schematic�representation�of�a�quantum-optical�EAP,�which�relies�on�a�challenge-response�mechanism.�The�enrolment�stage�is

performed�once�by�the�manufacturer,�while�the�verification�stage�takes�place�each�time�the�holder�of�the�PUK�has�to�be�authenticated.�

Electronic communications and transac-tions constitute one of the main pillars ofour society. The role of cryptography isto ensure the stability of this pillar, byproviding techniques for keeping infor-mation secure, for determining whetherinformation has been maliciouslyaltered, and for determining whoauthored pieces of information. To alarge extent, modern public-key cryp-tography relies on mathematical prob-lems, such as the factorisation of largeintegers and the discrete logarithm

problem, which are considered to behard to solve on classical computers, inthe sense that there are no known effi-cient classical algorithms for solvingthese problems for all integers in poly-nomial time. Given, however, that thenon-existence of such algorithms hasnever been rigorously proved, most ofthe widely used public-key cryptosys-tems are susceptible to advances inalgorithms or in hardware (computingpower). Indeed, Peter Shor has provedthat a quantum computer could solve

both of the aforementioned mathemat-ical problems efficiently.

Though a fully functional universalquantum computer of the necessary sizeto break widely used cryptosystems isstill far off in the future, there is a neces-sity for ensuring the security, theintegrity and the authenticity of datathat is encrypted or digitally signedtoday, and they have long lifetime.Unfortunately, the currently availableQKD systems become inefficient for

=L!

*+,-./0 1.

=.")0.0-1*+2*@2NO2340AA"+/" '"()*+("

;0-060("

340AA"+/" '"()*+("

M M

121314

5253

54

%0(".

5.&"0(?."5"+-

2,#+667",*-)*-."(0#"-*"#7&,)6

=L!

*+,-./0

6.

=.")0.0-1*+2*@2NO2340AA"+/"

340AA"+/" '"()*+("

M M

121314

5253

54

%0(".

5.&"0(?."5"+-

%"#&'&()*&+,-)*--80,9."(0#"-*"#7&,)6

5.

!"#$"#

6. 7 1.: P"(Q*

;(("1*<"="(*;"<1(1*+

=,Q

! *+,-/0R2>?0+-?52(-0-"2*@2A1/4-2)0.05"-".1S":26#202("-2*@2<A0((1<0A2)0.05"-".(25O2

! ;1@@"."+-2>?0+-?52<40AA"+/"(20."2+*+D*.-4*/*+0A ;"<1(1*+

large networks (many users), where theestablishment of many pairwise secretkeys is needed, and the key manage-ment remains a major problem.Moreover, QKD alone does not addressmany other cryptographic tasks andfunctions, which are of vital importancein everyday life (e.g., authentication,non-repudiation, integrity, etc.).

Cryptographic research in our theoret-ical group started in 2007, and itsemphasis is on the design and the devel-opment of quantum cryptographicprimitives and protocols, which rely onfundamental quantum-optical systemsand processes, and offer security againstquantum adversaries [1-3]. Typicalquantum-optical systems are singlephotons, light in various quantumstates, atoms, waveguides, and cavities.A description of our activities andrelated publications can be found at[L1] and [L2].

Entity authentication (identification) isan important cryptographic task, inwhich one party (the verifier) obtainsassurance that the identity of anotherparty (the claimant) is as declared,thereby preventing impersonation.Most of the entity authentication proto-cols (EAPs) used for everyday tasks(e.g., transactions in automatic tellermachines), rely on dynamic challenge-response mechanisms, which combinesomething that the claimant knows(e.g., a PIN), with something that theclaimant possesses (e.g., a smart card).In such mechanisms, after the user typesin the correct PIN, the smart card ischallenged with random numericalchallenges, and the verifier checks if theresponses of the card are valid.Conventional EAPs are not totallyimmune to card-cloning, while they aresusceptible to emulation attacks, inwhich an adversary knows the chal-lenge-response properties of the smartcard (e.g., by hacking the database ofchallenge-response pairs), and his taskis to intercept each numerical challengeduring the verification stage, and sendto the verifier the expected response.

Currently, optical physical unclonablekeys (PUKs) are considered to be themost promising candidates for thedevelopment of highly secure EAPs.Such PUKs are materialised by anoptical multiple-scattering disorderedmedium, and they are considered to beunclonable, in the sense that theircloning requires the exact positioning(on a nanometre scale) of millions ofscatterers with the exact size and shape,which is considered to be a formidablechallenge not only for current, but forfuture technologies as well. Typically, aPUK-based EAP relies on a challenge-response mechanism, in which the PUKis interrogated by light pulses with ran-domly chosen parameters, and accept-ance or rejection of the PUK is decidedupon whether the recorded responsesagree with the expected ones. Although,in general, PUK-based EAPs are morerobust against cloning than conven-tional EAPs, they are still vulnerable toemulation attacks when challenges per-tain to classical light, and the verifica-tion set-up is not tamper-resistant. Toeliminate this vulnerability, in collabo-ration with E. Diamanti (CNRS,Université Pierre et Marie Curie), wehave proposed a novel quantum-opticalEAP in which a PUK is interrogated byrandomly chosen non-orthogonalcoherent quantum states of light (seeFigure 1) [1]. The response of the PUKto a quantum state (challenge) is sensi-tive to the internal disorder of the PUK,which makes our protocol collisionresistant, and robust against cloning.Moreover, on-going research shows thatthe security of our protocol against anemulation attack relies on the laws ofquantum physics, which do not allowunambiguous discrimination betweennon-orthogonal quantum states, whileinformation gain cannot be obtainedwithout disturbing the quantum stateunder interrogation. The proposed pro-tocol can be implemented with currenttechnology, and its performance underrealistic conditions is the subject offuture theoretical and experimentalwork.

Links:

[L1] http://www.quantum-technology.gr/[L2] http://gate.iesl.forth.gr/~nikolg

References:

[1] G. M. Nikolopoulos, E. Diamanti:“Continuous-variable quantumauthentication of physicalunclonable keys”, ScientificReports, 2017, 7, p. 46047.

[2] G. M. Nikolopoulos, T. Brougham:“Decision and function problemsbased on boson sampling”,Physical Review A, 2016, 94, p.012315.

[3] G.M. Nikolopoulos: “Applicationsof single-qubit rotations inquantum public-key cryptography”,Physical Review A, 2008, 77, p.032348.

Please contact:

Georgios M. NikolopoulosIESL-FORTH, [email protected]




Quantum computing and quantum infor-mation in general offer incredible bene-fits to our information society. Since theoriginal discoveries of better than clas-sical security in quantum key distribu-tion and super-classical computationaladvantage, the field has exploded withnew possibilities for exploiting quantumencoding.

In quantum cryptography we have seennew protocols for coin flipping,quantum money and secure multipartycomputation offering functionality andor security that is not possible classi-cally. For certain communication tasks,quantum techniques provide an expo-nential gap between what is possiblequantumly and classically. Even beforeuniversal quantum computers are born,sub-universal devices will have aplethora of applications from quantumlearning to simulation. In quantummetrology, quantum sensing providesprecision in measurements that wouldsimply not be possible without uniquelyquantum features.

A remarkable family of multipartiteentangled states acts as genericresources for almost all these applica-tions. Graph states are described in oneto one correspondence with a simplegraph, where vertices represent qubits (atwo dimensional quantum system whichis the basic quantum information unit)and edges represent a particular entan-glement preparation. They are universalresources for quantum computation, actas codes for quantum error correctionand are the entangled resource for manycommunication and cryptographic pro-tocols. Perhaps their most compellingstrength is that they can be connected indifferent ways to allow the utility of onefunction to be combined with another ina natural way. In this sense, these “graphstates” are like quantum Lego – decidewhat you want to make and put them

together in the right way to achieve it.This capacity will be key in taking thebest advantage of quantum technologiesin future quantum networks. Indeed,typically, more sophisticated applica-tions are built up by combining basicprotocols.

The natural connection to graph theoryhas proven an additional benefit of thegraph state approach. It turns out thatone can understand many properties oftheir use for quantum information –how computation flows, where infor-mation sits – entirely in terms of theunderlying graph properties. Thisallows many graph theory techniques tobe put into play to push more whatquantum advantages can be had.Examples include graphical characteri-sation of where information sits insecret sharing [1] and the application ofrandom graph techniques for findingoptimal codes [2].

Another consequence of their ubiquityin quantum information is that there hasbeen a lot of effort to demonstrate themexperimentally. Indeed they representthe cutting edge in what entangledstates are prepared, with audaciousexperiments preparing graph states ofthousands of qubits. Experiments rou-tinely produce and control graph statesof up to 10 qubits in different media inoptics, atomics and ions.

There are now several groups acrossEurope and the world working onexploring graph state quantum informa-tion processing. In a series of workswith the groups of Mark Tame (Durban,South Africa) and John Rarity (Bristol,UK) we have been pushing the quantumLego aspect, in particular. In one exper-iment we demonstrated a graph stateprotocol, which combined three dif-ferent protocols to enable verified secretsharing [3]. This flexibility here proved

crucial – by combining protocols we getfunctionality that is better than could beachieved by any single protocol in iso-lation. But there are many excitingthings still to do, and we’re still figuringout how graph states can be used, inseveral directions to push the limits ofquantum information. For example,graph states have proven a fertile testspace to understand the resources ofnon-locality and contextuality and theirrole in many quantum advantages. Thegraphical notion of flow of informationalso lends itself to the study of excitingnew directions in quantum informationwhere the inherent ambiguity in causalorder has been shown to be yet anothersource of quantum advantage. Recently,we have also seen that graph states areresources for the generation of quantumrandomness, an almost generic resourcein quantum information and key in ourunderstanding in much of physics.

References:

[1] D. Markham, B. C. Sanders:“Graph states for quantum secretsharing”, Physical Review A 78.4(2008): 042309.[2] J. Javelle, M. Mehdi, S. Perdrix:“New Protocols and Lower Bounds forQuantum Secret Sharing with GraphStates”, TQC 12 (2012): 1-12.[3] B. A. Bell, et al.: “Experimentaldemonstration of a graph statequantum error-correction code”,Nature communications, 5, 2014

Please contact:

Damian Markham CNRS, LIP6, Sorbonne Université[email protected]

Quantum Lego: Graph States for Quantum

Computing and Information Processing

by Damian Markham (LIP6, CNRS - Sorbonne Université)

The massive global investment in quantum technologies promises unprecedented boosts for

security, computation, communication and sensing. In this article we explore the use of so-called

‘graph states’ – a family of multipartite entangled states which act as ubiquitous resources for

quantum information, are easily adapted for different tasks and applications, and can be combined

in ways that fuses different utilities.

Nowadays, it is theoretically predictedwith paper and pencil, and experimen-tally demonstrated in the lab, that thephysical state of most multi-objectquantum mechanical systems cannot bedescribed by looking only at their indi-vidual components, and displaysstronger-than-classical correlations. Invirtue of this point, manipulating someof the components has a global effect onthe whole system, even if the compo-nents are spatially separated – some-thing expressed by the currently over-used and somehow unwelcomingEinstein’s “spooky action at distance”.This phenomenon is known as “entan-glement”.

When talking about entanglement, theaverage mathematical person tends tonaturally refrain from alluding to thehalo of mysticism surrounding this term,but to focus on the intricate structure ofmatrices and operators in compositeHilbert spaces. However, while entan-glement may appear as a flat conse-quence of the tensor product used tocombine systems as in wave-mechanics,it also has the status of a legitimate andpossibly fundamental physical quantity.In truth, its freshly discovered applica-tions range from information-theoreti-cally secure solutions for distributingcryptographic keys to an assortment of

remarkable protocols for transmittinginformation, including superdensecoding and teleportation.

Two-player non-local games on graphshave been shown to be a particularlyfruitful arena for rigorous research intothe power of entanglement and, moregenerally, the correlations induced bythe axiomatic choices leading to theirdifferent physical theories. In the pro-posed framework, two players, Aliceand Bob, receive vertices of somegraphs. Their task is to respond to ques-tions without knowing each other’s ver-tices; answers need to satisfy a certainproperty as a function of the receivedvertices. Depending on the type of cor-relations displayed by the physicalworld of Alice and Bob, this situationsuggests a collection of new graphparameters, which are captured by richmathematical structures, and give anoccasion to generalise graph theoryideas to functional analysis. One ofthese quantities is now called quantumchromatic number and it is known to beloosely upper bounded by the morefamiliar chromatic number – when, inthe non-local game, answers need to bedifferent for adjacent and equal foridentical vertices. The quantum chro-matic number was the first quantumgraph parameter to be studied. Such

research direction has eventually rami-fied into multiple routes. A briefaccount of the salient points follows.

In 1956, Shannon defined the zero-errorcapacity of a communication channel asthe largest rate at which information canbe transmitted over the channel witherror probability zero. The notion hascontributed to fuel a great amount ofresearch in semidefinite programmingand structural graph theory. Berge’s per-fect graphs were remarkably motivatedby the zero-error capacity and theLovász theta function was introduced asan upper bound. When the parties canshare and locally manipulate entangledstates, the analogue capacity has alsobeen proved to be bounded by Lovásztheta, but to be exponentially larger invarious single-shot and asymptoticcases [1, 2]. Given the link betweenzero-error capacity and the problem oftransmitting data over a broadcastchannel (e.g., coaxial cable or satellite),this result highlights a prospectivequantum advantage exploitable in real-world communications.

In 1976, Connes casually formulated aconjecture about a fundamental approx-imation property for finite vonNeumann algebras – the Connesembedding problem. Over time, many



Graph Parameters and Physical Correlations:

from Shannon to Connes, via Lovász and tsirelson

by Simone Severini (University College London)

Quantum information theory builds bridges between combinatorics, optimisation, and functional analysis.

Figure�1:�Two�graphs�on�24�vertices�that�are�quantum�isomorphic�but�not�isomorphic.�The�construction�is�related�to�the�FGLSS�reduction�from

inapproximability�literature,�as�well�as�the�CFI�construction.

unexpected equivalent statements forthe conjecture have emerged. A hier-archy beyond the quantum chromaticnumber has led to a novel reformulationof the Connes conjecture, and generateda fresh effort towards its solution [L1].The new approach is centred on combi-natorial ideas lifted to the realm of oper-ator algebras. Moreover, extending thismathematical landscape, hierarchies ofquantum graph parameters associated tocorrelations can be placed in the frame-work of (tracial noncommutative) poly-nomial optimisation [L2]

In 1993, Tsirelson proposed some prob-lems concerned with deciding whetherthe axiomatic mathematical models ofnon-relativistic quantum mechanics,where observers have operators actingon a finite dimensional tensor productspace, and algebraic quantum fieldtheory, where observers have com-muting operators on a (possibly infinitedimensional) single space, produce thesame set of correlations. A 2016 break-through, based on geometric grouptheory, settled the “middle” Tsirelsonproblem, by observing that the set of(tensor-product) quantum correlationsis not closed [L3]. Interestingly, a suc-cessive alternative proof of this resultmakes use of quantum graph parameters[L4].

This new technical machinery becamefurther consolidated via a quantum ver-sion of graph homomorphism, a

familiar but powerful generalisation ofgraph colouring. This new technicalmachinery became further consolidatedvia a quantum version of graph homo-morphism, a familiar but powerful gen-eralisation of graph colouring. The newtype of homomorphism suggestedrelaxations of graph isomorphism tosettings corresponding to various phys-ical theories. The findings of this line ofresearch are surprising. While fractionalisomorphism corresponds to sharingsome type of correlations stronger thanentanglement – hence, it gets an opera-tional interpretation, – quantum isomor-phism is obtained by relaxing theinteger programming formulations forstandard isomorphism to Hermitianvariables [3] (see Figure 1).

The QCIAO Collective is an interna-tional collaboration focused on thetopics of this article [L5].

Links:

[L1] https://arxiv.org/abs/1503.07207 [L2] https://arxiv.org/abs/1708.09696[L3] https://arxiv.org/abs/1606.03140 [L4] https://arxiv.org/abs/1709.05032 [L5] http://www.qciao.org/

References:

[1] R. Duan, S. Severini, A. Winter:“Zero-error communication viaquantum channels, non-commutative graphs and a quantumLovasz theta function”, IEEETrans. Inf. Theory 59(2):1164-1174, 2013.

[2] J. Briët, H. Buhrman, D. Gijswijt:“Violating the Shannon capacity ofmetric graphs with entanglement”,PNAS 2013 110 (48) 19227-19232.

[3] A. Atserias, et al.: “Quantum andnon-signalling graphisomorphisms”, ICALP 2017.

Please contact:

Simone Severini University College London, UK+44 (0)20 3108 [email protected]


Some�members�of�the

Quantum�Computing,

Information,�and

Algebras�of�Operators

(QCIAO)�collective�

(lecturers�of�the�LMS

Research�School

Combinatorics�and

Operators�in�Quantum

Information�Theory,

Belfast�September�2016).



1

2

Figure�1:�Frequency�correlation�plot�(joint�spectral�intensity)�between�the�photons�of�the

generated�pair.�The�nearly�circular�distribution�indicates�a�low�degree�of�correlations;�in

contrast�a�more�elongated�spectral�distribution�would�point�to�a�low�purity�of�the�photons.

Although the current favourites for ascalable quantum computer seem to besuper-conducting qubit implementations,a pure photonic-based solution shouldnot be disregarded, since it offers roomtemperature operation and small foot-prints. As the standard approach toquantum computing is based on thequantum circuit model, a generalisationof the classical circuit, featuring logicaloperations such as AND and XOR gates,single-qubit gates (e.g. rotations) andtwo-qubit gates (e.g. controlled-NOT)are required for universal quantum com-puting. This quantum circuit model is notvery well suited to a photonic implemen-tation, since the photon-photon interac-tion is negligible and a two-photon gateis hence not feasible.

Computations over the quantuminternetWith the near advent of quantum net-works, consisting of distribution ofsingle or entangled photons betweenusers, a photonic quantum computerwould have another advantage. Unlikematter based quantum computers, whichrequire a photon/matter quantum inter-face, a cluster-state quantum computercould be directly connected to a global“quantum internet”. Protocols whichrely on a quantum link between user andquantum computer exist today. Imaginethe scenario, in which quantum com-puters are still very large and expensive,and clients can only request computationtime from a remotely located quantumcomputer centre which is not necessarilytrustworthy. Using the blind quantumcomputing protocol [1], the client’sinputs, outputs and computation remainperfectly private, even to the operatorsof the quantum computer centre. Inanother algorithm, quantum enabledone-time programing, the program itselfis encoded onto photons, sent to a clas-sical processing unit and executed.Since the qubits are destroyed during themeasurement process, the program willonly run once, a much sought-after prop-

erty for cryptographic application andsecure software distribution.

Cluster state computingA scalable model for photonic quantumcomputation is the so called one-way ormeasurement based quantum computer[2]. In this approach, the actual compu-tation proceeds by single qubit meas-urements only, on a large entangledresource state (cluster state). In a pho-tonic implementation, the cluster stateis a set of large numbers of photons,which are entangled to each nearestneighbour. The advantage lies in thefact, that there is no difference betweenperforming a single-qubit or two-qubitgate operation. In both cases, there willbe only measurements on single pho-tons, a feat easy to achieve. The main challenge lies of course in thecreation of the entangled cluster statewhich sounds like a taunting task, espe-cially for large states of several hun-dreds of qubits. Fortunately, it can beshown that heralded photon sources suf-fice to build a cluster state of any size.Of course efficiencies have to beincreased and losses minimised, but in

general, multiplexing a large number ofsuch sources together will produce thenecessary entangled state. One keyissue is that the photons produced mustbe indistinguishable, i.e. they must lookall the same. Therefore, no correlationsin frequency, time or spatial modes areallowed between the photons. Thefigure of merit for this property is thepurity, photons that are completelyuncorrelated have a high purity whereasphotons which correlate in a degree offreedom show a low purity. Figure 1illustrates the absence of frequency cor-relation between two generated singlephotons.

The GHz entanglement sourceThe AIT-Austrian Institute ofTechnology in Vienna successfullydeveloped a source for polarisation-entangled photon pairs which can beoperated at a tunable repetition rate ofup to 40 GHz to be used in cluster stategeneration [3]. Our source is based on aSagnac-interferometer, where a non-linear optical medium is pumped by twopulsed laser beams from two diametri-cally opposed directions. By virtue of a

High-Speed Entanglement Sources

for Photonic Quantum Computers

by Fabian Laudenbach, Sophie Zeiger, Bernhard Schrenk and Hannes Hübel (AIT)

Photonic quantum computers promise compact, user-friendly packaging. The building blocks of

such an implementation comprise of sources for efficient production of photons with high purity.

To increase the clock speed of the computation, such sources need to operate in the GHz range.

nonlinear process called parametricdownconversion (PDC), some of thelaser photons would decay within thecrystal and give birth to two photonswith half the energy each. Entanglementemerges from the lost “which-wayinformation” after recombination of thetwo paths: There is no possible way totell which pump beam produced thetwin photons; therefore the two possibleorigins exist at the same time in aquantum superposition, as shown in

Figure 2. The photons are generatedwith high purity and are highly entan-gled (95%).

OutlookIn the next step, we will simultaneouslypump three Sagnac-interferometers(instead of one) in order to generatelarger cluster states. Moreover, we willreplace our avalanche photo diodes bysuperconducting nanowire detectorswhich not only have significantly

higher detection efficiency but alsooperate at a higher time resolution. Thiswill allow us to generate cluster statesclose to the maximal pulse rate of thelaser at 40 GHz. While enlarging thenumber of photons in the cluster state,we will also strive to reduce the overallsize of the experiment using photonicintegration techniques.

AcknowledgementsThis work was funded by the AustrianResearch Promotion Agency (Österre-ichische Forschungsförderungsgesell-schaft, FFG) through the project KVQ(No. 4642983).

Link:

https://www.ait.ac.at/themen/optical-quantum-technologies

References:

[1] S. Barz et al.: “Demonstration ofBlind Quantum Computing”,Science 335 (2012), 303

[2] R. Raussendorf and H. J. Briegel:“A One-Way Quantum Compute”,PRL 86 (2001) 5188

[3] F. Laudenbach et al.: “Modellingparametric down-conversionyielding spectrally pure photonpairs”, Opt. Exp. 24 (2016), 2712

Please contact:

Hannes Hübel, Martin StierleAIT Austrian Institute of Technology,[email protected],[email protected]


Figure�2:�Left:�Principle�of�entanglement�generation.�Depending�on�the�path�of�the�pump�laser

(red),�the�vertical�(green)�and�horizontal�(blue)�photons�are�exiting�at�different�ports;�Right:

Experimental�setup�of�Sagnac�interferometer,�the�red�pump�beam�enters�from�the�right�side.

Between 2011 and 2013, I had theopportunity to collaborate on the Pan-American project QCS (QuantumComputer Science) devoted to theimplementation and analysis of quan-tum algorithms. Researchers from dif-ferent areas, ranging from physicists totheoretical computer scientists, wereworking on this project; I was at the lat-ter end of this spectrum. We were givena set of seven, state-of-the-art quantumalgorithms instantiated on some con-

crete problems, with the task to imple-ment and estimate the resources neces-sary to effectively run them, in theevent of an actual quantum computer.To this end we developed the quantumprogramming language Quipper [L1].

Quantum algorithms are not usuallydesigned as practical handles but insteadas tools to explore complexity bound-aries between classical and quantumcomputation: for a given problem, as the

size of the input parameter grows, canwe asymptotically go faster with the useof a quantum memory than with purelyclassical means? It turns out that manyinteresting problems have this property:many fields ranging from big-data tochemistry and pharmaceutic could ben-efit from the use of quantum algorithms.The natural question is how to con-cretely implement these quantum algo-rithms, estimate the required resourcesand validate the implementation.

A formal Analysis of Quantum Algorithms

by Benoît Valiron (LRI – CentraleSupelec, Univ. Paris Saclay)

Moving from the textual description of an appealing algorithm to an actual implementation

reveals hidden difficulties.

In the realm of classical computation,implementing an algorithm implies thechoice of a programming language tocode it, a platform to run the resultingprogram, and a compiler that can turnthe code into a program executable onthis platform. The realm of quantumcomputation is currently less developed:several competing platform co-exists,and in 2011, at the beginning of the QCSproject, no scalable quantum program-ming language even existed.

From a programmer’s perspective,quantum computation is very close toclassical computation. The main differ-ence lies in the use of a special kind ofmemory with exotic properties: in par-ticular, quantum data is non-duplicableand reading quantum data is a proba-bilistic operation. The interaction withthe quantum memory is done by asequence of elementary operations thatare summarised by what is known as aquantum circuit. A quantum algorithmmainly consists of the construction ofsuch circuits, their execution on thequantum memory, and various classicalpre- and post-processing.

This hints at the main required designchoice: a quantum programming lan-guage is primarily a circuit-descriptionlanguage. However, quantum algorithmsare not simply fixed, static quantum cir-cuits. Unlike classical circuitry such asFGPAs, quantum algorithms describefamilies of quantum circuits: the circuitdepends on the size and shape of theinput data. A quantum programming lan-guage therefore has to account for thisparametricity. Quipper has beendesigned from the ground up with theseaspects in mind. Following a successfultrend in domain specific languages,

Quipper is an embedded language withina host language. The chosen host lan-guage is Haskell. This modern, func-tional language features an expressivetype system making it easy to define andenforce the specificities of circuit con-struction and manipulation.Parametricity is naturally obtained withthe use of lists and list combinators. Theconstruction of a circuit is modelled asprinting on a particular kind of outputchannel: generating an elementaryinstruction corresponds to writing it onthe channel. Within Haskell, this kind ofside-effect can naturally be encapsulatedwithin a type construct known as monad.This automatically outlaws various ill-defined programs, renders their codingless error-prone while easing debugging.

Quipper has been used to code largealgorithms and perform logical resourceestimation: for a given set of parametersto the problem, what is the size of circuitgenerated by the algorithm? The size canbe counted in the number of elementary,logical operations and in the size of therequired quantum memory footprint.The analysis shows that for naive imple-mentations, these numbers can be quitelarge, yielding unrealistic circuits. Thisanalysis tends to show that producingusable algorithms requires more thanconcentrating on complexity analysis.Nowadays, it is possible to do so: withthe help of programming languages suchas Quipper, a quantum algorithmdesigner can effectively analyse and tuneits algorithm for concrete use-cases.

The advent of modern quantum pro-gramming languages clears a pathtowards the design of quantum compila-tion stacks and tools for certification ofquantum programs. Quipper is currently

the backbone of two projects aiming atsuch goals. The European Itea3 projectQuantex spanning across France,Netherlands and Germany gathersAtos/Bull, CEA/Leti, LORIA, LRI,TUDelft, KPN, Siemens and Univ.Tübingen. Quantex aims at developingthe programming environment aroundQuipper and focuses on the compilationtowards emulation of quantum compu-tation. The recently started French ANRSoftQPro is centered on a formalisationof Quipper’s semantics toward certifica-tion, and on the development of a com-pilation stack based on the graphical cal-culus ZX, envisioned as a more naturalintermediate representation than exist-ing proposals.

Link: [L1] https://kwz.me/hB6

References:

[1] A. S. Green, P. L. Lumsdaine, N. J.Ross et al.: “Quipper: A ScalableQuantum Programming Language”,in Proc. of PLDI 2013, ACMSIGPLAN Notices, Vol. 48 Issue 6,pp. 333-342, 2013.

[2] B. Valiron, N. J. Ross, P. Selinger etal.: “Programming the QuantumFuture”, Communications of theACM, Vol. 58 No. 8, pp. 52-61,2015.

[3] A. Scherer, et al.: “Concreteresource analysis of the quantumlinear-system algorithm used tocompute the electromagneticscattering cross section of a 2Dtarget”, Quantum InformationProcessing 16:60, 2017.

Please contact:

Benoît Valiron, LRI – CentraleSupelec,Univ. Paris Saclay, [email protected]



Quantum computers have the ability toreshape the modern world, offering hugecomputational speed-ups for a widevariety of problems in mathematics,

physics, chemistry, and computer sci-ence. However, the existing andemerging quantum computationaldevices face stringent limitations in

memory, computational power, and tol-erance to noise, making it crucial todevelop sophisticated techniques foroptimising the software which drives

diagram transformations Give a New Handle

on Quantum Circuits and foundations

by Aleks Kissinger (Radboud University)

String diagrams provide a powerful tool for uncovering hidden algebraic structure in quantum

processes. This structure can be exploited to optimise quantum circuits, derive fault-tolerant

computations, and even probe the foundations of physics.


these systems. Quantum circuits havebecome the de facto “assembly lan-guage” for quantum software. Theydescribe computations in terms of aseries of primitive operations, calledquantum gates, performed on a registerof quantum bits (qubits), which is thenmeasured to give a result. Whilequantum gates are useful as buildingblocks for computations, they lack awell-understood algebraic structure,making it difficult to understand whentwo circuits are equivalent (i.e., describethe same computation) or to transformone circuit into another for the sake ofoptimisation or fault-tolerance.

In 2008, researchers at OxfordUniversity produced a unique solutionto this problem: the ZX-calculus.Originally developed as an abstractmethod for studying the behaviour ofcomplementary observables in quantummechanics (indeed the ‘Z’ and ‘X’ referto the complementary Pauli observablesof the same name), the ZX-calculusquickly showed itself to be a useful

practical language for reasoning aboutquantum circuits, as well as other qubit-based models of computation, such asmeasurement-based quantum computa-tion. The ZX-calculus works by decom-posing quantum gates into even moreprimitive components, called “spiders”(Figure 1). These basic pieces satisfy asmall number of algebraic laws, whichin turn yield a great deal of power. Forexample, the four laws shown in Figure2 suffice to transform any two equiva-lent Clifford quantum circuits into thesame circuit. Clifford circuits are awell-studied class of circuits which canbe simulated efficiently on a classicalcomputer, so it may not be too sur-prising that the ZX-calculus gives aneasy, algebraic handle on circuit equiva-lence. However, what is surprising isthat groups in Oxford and LORIA (aresearch unit affiliated with ERCIMmember Inria) have shown in recentmonths that an extension to these rulescan decide equality for Clifford+T cir-cuits [1] and even a fully universalfamily of quantum circuits [2]. These

families of circuits are capable of pro-ducing any quantum computation imag-inable (or, in the case of Clifford+T,approximating it to arbitrarily high pre-cision). Thus, a complete algebraiccharacterisation of equivalence forthese circuits is a major breakthrough.

The ZX-calculus—including the exten-sions proposed by the Oxford andLORIA groups—is based on string dia-grams, which can be seen as a sort ofgeneralisation of quantum circuits.They first appeared in the work ofPenrose in the 1970s, and since thenhave been applied to a broad range ofapplications in physics and computerscience, including tensor networks inhigh-energy physics, signal-flowgraphs, electrical and electronic cir-cuits, computational linguistics, andconcurrent computation. Earlier thisyear, Coecke and Kissinger published atextbook which gives a comprehensiveintroduction to quantum theory,quantum computation, and quantumfoundations purely in the language ofstring diagrams [3].

String diagrams not only provide aunique and intuitive way to introducethe core concepts of quantum theory,but also a dramatically different way ofworking with quantum mechanicalprocesses. Within the diagrammaticapproach to quantum theory (a.k.a.“Categorical Quantum Mechanics”, seelink [L2]), concepts such as connec-tivity, composition, and interaction takecentre stage, whereas concrete Hilbert-space calculations are secondary. Forexample, foundational questions around

Figure�2:�The�rules�of�the�ZX-calculus,�which�suffice�to�derive�all�equations�that�hold�between�Clifford�circuits.�

Figure1:�Decomposing�quantum�gates�into�more�primitive�pieces�called�“spiders”.



quantum non-locality and quantumcausal structure can be posed in terms ofwhether a diagram decomposes in a cer-tain way across time and space, andwhat consequences that decompositionhas on our observations.

More pragmatically, quantum algo-rithms and communication protocolscan be proven correct using diagramtransformation rules, even in casesinvolving far too many qubits for con-crete calculation. To assist with pro-ducing, checking, and sharing theseproofs, researchers at RadboudUniversity, University of Strathclyde,and Oxford have developed a tool calledQuantomatic (Figure 3). Quantomatic isa “proof assistant for diagrammatic rea-soning”. Using a combination of human-guided diagram transformations andautomated rewrite strategies (program-mable in a Python-based strategy lan-guage), it is possible to perform a varietyof tasks in Quantomatic, such as opti-mising small to medium-sized quantumcircuits, verifying multi-party communi-cation protocols, and computingencoded logical operations within cer-tain families of quantum error correctingcodes.

The teams in Nijmegen, Oxford,LORIA, and Strathclyde, with the helpof collaborators at LRI, Durham, andthe UK’s NQIT Quantum Hub, are nowaiming to produce fully automated tech-niques for circuit optimisation, scale upto large computations on hundreds oflogical qubits (with error correction),and develop new kinds of transforma-tion procedures to work within the con-straints of first-generation quantumhardware, such as limited topologies forqubit interactions and distinguishedgate- vs. memory-optimised physicalqubits.

Links:

[L1] http://quantomatic.github.io[L2] http://cqm.wikidot.com[L3] https://kwz.me/hB7

References :

[1] E. Jeandel, S. Perdrix, R. Vilmart:“A Complete Axiomatisation of theZX-Calculus for Clifford+TQuantum Mechanics”, Preprint,arXiv:1705.11151. 2017.

[2] K. Ng, Q. Wang: “A universalcompletion of the ZX-calculus”,Preprint, arXiv:1706.09877. 2017.

[3] B. Coecke, A. Kissinger: “PicturingQuantum Processes: A First Coursein Quantum Theory andDiagrammatic Reasoning”,Cambridge University Press. 2017.

Please contact:

Aleks KissingeriCIS Radboud Universiteit, [email protected]

Figure�3:�Quantomatic,�a�proof�assistant�for�diagrammatic�reasoning.�Here,�it�is�computing�a�transversal�implementation�of�a�CCZ�gate�within�a

quantum�error�correcting�code�(namely,�the�[[8,3,2]]�colour�code;�see�links�[L1]�and�[L3]�for�more�details).


Quantum computers are expected tosolve problems which are intractable toclassical information processing. Builton the principles of quantum mechanics,they exploit the complex and fascinatinglaws of nature at the molecular andatomic level, which usually remainhidden from view. By harnessing thisquantum behaviour, quantum computingcan run new types of algorithms toprocess information in extremely largestate spaces. These algorithms may oneday lead to revolutionary breakthroughsin materials simulation, the optimisationof complex manmade systems, and inmachine learning.

Quantum computers are based on qubits,which operate according to two keyprinciples of quantum physics: superpo-sition and entanglement. Superpositionmeans that each qubit can represent botha “1” and a “0” at the same time [1]. Itleads to what is called “quantum paral-lelism”, an effect that allows an n-qubitregister to store all 2n possible states atthe same time. Entanglement means thatthe state of multiple qubits can be corre-lated with each other; that is, the state ofone qubit (whether it is a “1” or a “0”)depends on the state of one or moreother qubits and the qubit can no longerbe described separately. Using the prin-ciple of entanglement creates additionalprocessing capabilities which are notavailable on classical processors.

IBM’s quantum devices are built onsuperconducting Josephson junctiontechnology which requires cryogenictemperatures as depicted in Figure 1.The quantum bits are anharmonic LCresonators, known as transmon qubits[2], where the non-linear inductor (L) isimplemented by a Josephson junction.The left outer device in Figure 2 showsfive square shaped qubits (a). The mean-dering connections (b) are coplanarwaveguides (CPW). They are used tocouple qubits with each other for infor-

in Figure 2 (left and middle). Currently,5-qubit and 16-qubit devices are pub-licly available to run user programsfrom the cloud. Alternatively, a cloud-based simulator software which handlesup to 20 qubits can be chosen.Examples of the three different pro-gramming interfaces are depicted inFigure 3. The graphical user interface isthe simplest way to access the real

Quantum Computers and their Software

Interfaces

by Peter Mueller, Andreas Fuhrer and Stefan Filipp (IBM Research – Zurich)

Scientific groups in industry and academia have made enormous progress in the implementation of

first quantum computer prototypes. IBM’s quantum experience with five and 16 qubits are already

publicly accessible in the cloud. Three “standard” software interfaces are available. Client systems

with 20 qubits ready for use and the next-generation IBM Q system is in development with the first

working 50 qubit processor.

mation processing and to couple qubitsto I/O ports (c) to manipulate, write andread the qubits. All the qubit manipula-tions and readout processes are con-trolled by microwave pulses in therange of 4 to 8 GHz.

The “IBM Q experience” provides aplatform to explore basic quantum cir-cuits on real quantum devices as shown

Figure�1:�Open�cryogenic

system�for�a�50�qubit

quantum�computing�device.

The�device�is�hidden�inside

the�shielding�tube�attached�at

bottom�center.�The�copper

structures�to�the�left�and

right�of�the�shielding�tube

contain�parametric�quantum

amplifiers�to�readout

information�from�the�qubits.

Figure�2:�An�experimental�5-qubit�(left)�and�16-qubit�(mid)�device�as�used�in�the�publicly

available�“IBM�Q�experience”�and�a�package�as�used�for�the�upcoming�20/50�qubit�devices

(right).�For�further�details�see�text.

quantum hardware using a webbrowser. OpenQASM [3] is a low levelhardware interface which enables com-patible software stacks to be built. Formore advanced developments, theQuantum Information Software Kit(QISKit) is an open-access programinginterface which provides highest func-tionality for working with both, the realquantum processors and various soft-ware based quantum simulators. The

open-access interface allows a user-application to integrate the quantumprocessor as an accelerator, through thecloud.

On the application side, calculations ofe.g., the ground state energies of smallmolecules such as H2, LiH and BeH2have been reported. Further examplesare shown in the links given below. The“IBM Q experience” provides a plat-

form to explore basic quantum circuitson real quantum processors based onsuperconducting qubits. It has beenused by more than 60,000 users and hasalready generated more than 35 third-party research publications on applica-tions of quantum computing. On thesoftware side, a growing number ofdevelopers are using QISKit. To sum-marise, the time is now to join forces,execute and continue to make keybreakthroughs on this revolutionaryjourney of information technology.

Links:

http://ibm.com/ibmq/http://www.qiskit.org/http://math.nist.gov/quantum/zoo/

References:

[1] M. A. Nielsen and I. L. Chuang,“Quantum Computation andQuantum Information”, 10th edt.,Cambridge University Press, 2010.

[2] J. Koch et al., “Charge-insensitivequbit design derived from theCooper pair box”, Phys. Rev. A 76,042319, 2007.

[3] A. W. Cross et al., “Open QuantumAssembly Language”,[arXiv:1707.03429], 2017.

Please contact:

Peter MuellerIBM Research – Zurich, [email protected]



Figure�3:�The�three�“IBM�Q�experience”�web�interfaces�–�graphical�user�interface�(top),

OpenQASM�assembler�(bottom�left)�and�the�Quantum�Information�Software�Kit�(bottom�right)�–

give�access�to�the�experimental�five�and�16�qubit�circuits.

Chevalley-Warning theorem in Quantum

Computing

by Gábor Ivanyos and Lajos Rónyai (MTA SZTAKI, Budapest)

Effective versions of some relaxed instances of the Chevalley-Warning Theorem may lead to

efficient quantum algorithms for problems of key practical importance such as discrete logarithm

or graph isomorphism.

The Theory of Computing ResearchGroup of the Informatics Laboratory atMTA SZTAKI has expertise in algebraicaspects of quantum computing,including quantum algorithms for alge-braic and arithmetical problems, as wellas application of algebraic methods asingredients of quantum algorithms.Some of our projects aim at discoveringhidden algebraic structures, e.g., symme-tries of certain objects. A main exampleis the so-called “hidden subgroupproblem”, which includes such promi-

nent special cases as the task of com-puting discrete logarithms and the ques-tion of finding isomorphisms of graphs.The object we are given is a function f

defined on a large finite group G and weare looking for the subgroup H con-sisting of all elements h for which f(xh) =f(x) for every x from G. In other words, His the group of elements whose actionleaves f invariant. (We remark note thatin most cases, we further require that f issuch that f(x) = f(y) if and only if y = xh

for some h from H.) Perhaps the simplest

and best known example of this isfinding periods for functions defined onthe integers. One of the greatest suc-cesses of quantum algorithms, Shor’smethod for factoring integers, is basedon finding such a period. Computing dis-crete logarithms in various settings are isalso an instance of the hidden subgroupproblem over abelian groups.

The graph isomorphism problem can becast as an instance of the hidden sub-group problem over a noncommutative


group G. In contrast with the commuta-tive case, for which efficient quantumalgorithms are known, the complexity ofthe noncommutative hidden subgroupproblem has remained open even forcertain groups that are very close tocommutative ones. Among the few posi-tive results in this direction, we mentionour polynomial time algorithm, devel-oped in a joint work [1], which findshidden subgroups in a fairly wide classof groups in which the order of the ele-ments is bounded by a constant. Theoverall progress is much more modesteven in “so-called two-step solvablegroups” (these are in a certain sensecomposed of two commutative groups)in which elements of larger order arepresent.

With our collaborators we found [2] thatthe hidden subgroup problem for a sub-class of such groups can be further gen-eralised to another class of problemsregarding hidden algebraic structure. Inthis class, the hidden object is a polyno-mial map between vector spaces over afinite field. Certain hidden subgroupproblems can be formulated as hiddenpolynomial map instances (there is areduction in the other direction as well,but this results in a bigger hidden sub-group problem). A simple illustrativeexample of a hidden polynomial map isas follows: let f(X) be an unknown uni-variate polynomial of constant degree.We have access to a quantum oraclewhich returns E(Y2-f(X)) for given pairs(X,Y). Here E is an unknown injectiveencoding of the field. The task is todetermine f (up to constant term). Wedeveloped [2,3] a polynomial timequantum algorithm for finding suchhidden polynomial maps under theassumption that they have constantdegree.

One of the critical ingredients of ourquantum algorithm is a classical algo-rithm that under certain conditions findsa nontrivial solution of a system of poly-nomial equations of a very special kind,for which the basic and famousChevalley-Warning theorem of numbertheory ensures the existence of a non-trivial solution. Our system is obtainedfrom a system of homogeneous linearequations by replacing each variable byits d-th power where d is a fixed positiveinteger:

The condition that allowed us a methodrunning in polynomial time is that thenumber of variables, compared to thenumber of equations and the degree d, issufficiently large. (Here by polynomialtime we mean time bounded by a func-tion polynomial in the bit size of thearray of the coefficients, which is thenumber of equations, m , times thenumber of variables, n, times the loga-rithm of the size of the base field). In thequantum setting in which our algorithmis applied, the degree is essentially thedegree of the hidden polynomial mapand the number of equations is related tothe dimension of the underlying spaces,while we are allowed to choose thenumber of variables. (Note however,that the system is not required to besparse, the n times m array of the coeffi-cients can be arbitrary.)

Observe that without any assumption onthe number n of variables, already overthe field consisting of three elements thequadratic case of the problem becomesNP-hard. This can be shown by a modi-fication of the standard reduction of SAT

to Subset sum. (In fact, that case of theproblem is just finding a zero-one solu-tion of the corresponding linear system.)On the other hand, from the Chevalley-Warning Theorem it follows that if n >md, then our system always has a non-trivial solution. Then an interestingquestion arises: how hard is it to find anontrivial solution? There is some evi-dence (such as the above mentionedhardness result) that this question is toodifficult when the number of variables isclose to the Chevalley-Warning boundmd. For this reason, we look for an effi-cient solution for relaxations in which nis substantially larger than this bound.First, it is worth noting that by a simpleand natural recursive algorithm, it iseasy to find a solution in polynomialtime, when the number n of variables isgreater than a function like d raised tothe m-th power. This method is usefulwhen m is constant. What can we saywhen d is kept constant? For this variantof the problem, we have developed amuch more sophisticated algorithm [3].This result is probably not optimal andan improvement could be a first steptoward quantum algorithms for findinghidden polynomial maps of higher

degree and toward hidden subgroupalgorithms in some more complexgroups.

References:

[1] K. Friedl, et al: “Hidden translationand translating coset in quantumcomputing” SIAM Journal onComputing 43 (2014), pp. 1-24.

[2] T. Decker, et al.: “Polynomial timequantum algorithms for certainbivariate hidden polynomialproblems”, Quantum Informationand Computation 14 (2014), pp.790-806.

[3] G. Ivanyos, M. Santha: “Solvingsystems of diagonal polynomialequations over finite fields,”Theoretical Computer Science 657(2017), pp. 73-85.

Please contact:

Gábor Ivanyos and Lajos RónyaiMTA-SZTAKI, [email protected],[email protected]

The major issue for building a quantumcomputer is thus to protect quantuminformation from errors. The quantumcounterpart of error correcting codesprovides an algorithmic solution, basedon redundant encoding, towards scalingup the precision in this context.However, it first requires technology toreach a threshold at which the hardwareachieves the following on its own:• all operations, in any big system, must

already be accurate to a very high pre-cision;

• the remaining errors follow a particu-lar scaling model, e.g., they are allindependent in a network of redun-dant qubits.

Under these conditions, adding morephysical degrees of freedom to encodethe same quantum information keepsleading to better protection. Otherwise,

the fact that new degrees of freedomand additional operations also carrynew possibilities for inducing errors orcross-correlations, can degrade theoverall performance.

In the QUANTIC lab at Inria Paris [L1],we are pursuing a systems engineeringapproach to tackle this issue by drawinginspiration from both mathematicalcontrol theory and the algorithmic errorcorrection approach (Figure 1).

A first focus of the group is to design ahardware basis where a large Hilbertspace for redundant encoding of infor-mation comes with just a few specificdecoherence channels. Concentratingefforts on rejecting these dominant noisesources allows more tailored and simplestabilising control schemes to be appliedto improve the quality of these building

blocks towards more complex error cor-rection strategies. Such hardware effi-ciency is typically achieved when manyenergy levels are associated to one andthe same physical degree of freedom. Aprototypical example is the “cat qubit”,where information about a single qubit isredundantly encoded in the infinite-dimensional state space of a harmonicoscillator, in such a way that the domi-nant errors reduce to the single photonloss channel; and this error is both infor-mation-preserving and non-destructivelydetectable. This overcomes the need fortracking multiple local error syndromesand applying according corrections in acoupled way, as would be the case for alogical qubit encoded on a network ofspin-1/2 subsystems. An implementationof this principle in superconducting cir-cuits has recently achieved the firstquantum memory improvement thanksto active error correction.

A second point of attention is the speedof operations. Indeed, quantum opera-tions take place at extremely fasttimescales — e.g., tens of nanosecondsin the superconducting circuits favouredby several leading research groups. Thisleaves little computation time for a dig-ital controller acting on the system.Conversely, it hints at the fact that asmartly designed quantum system couldstabilise fast on its own onto a protectedsubspace, possibly in combination witha few basic control primitives. Suchautonomous stabilisation can beapproached by a “reservoir engineering”strategy, where conditions are set up fordissipation channels to systematicallypush the system into a desired direction.This can enact e.g., fast reset of qubits,or error decoding and correction, all inone fast and continuously operating(“pumping”) hardware without anyalgorithmic gates. Such “reservoir engi-neering” also plays an important role ineffectively isolating quantum subsys-



Hardware Shortcuts for Robust Quantum Computing

by Alain Sarlette (Inria) and Pierre Rouchon (MINES ParisTech)

Despite an improved understanding of the potential benefits of quantum information processing and the

tremendous progress in witnessing detailed quantum effects in recent decades, no experimental team to date

has been able to achieve even a few logical qubits with logical single and two qubit gates of tunably high

fidelity. This fundamental task at the interface between software and hardware solutions is now addressed

with a novel approach in an integrated interdisciplinary effort by physicists and control theorists. The challenge

is to protect the fragile and never fully accessible quantum information against various decoherence channels.

Furthermore, the gates enacting computational operations on a qubit do not reduce to binary swaps, requiring

precise control of operations over a range of values.

Figure�1:�General�scheme�of�our�quantum�error�correction�approach�via�“hardware�shortcuts”.

The�target�“quantum�machine”�is�first�embedded�into�a�high-dimensional�system�that�may�have

strong�decoherence,�but�only�along�a�few�dominating�channels�(red).�Then,�a�part�of�the

environment�is�specifically�engineered�and�coupled�to�it�in�order�to�induce�a�strong�information-

stabilising�dissipation�(blue).�This�entails�stabilising�the�target�quantum�machine�into�a�target

subspace,�e.g.,�“four-legged�cats”�of�a�harmonic�oscillator�mode;�and�possibly,�replacing�the

measurement�of�error�syndromes�and�conditional�action,�by�a�continuous�stabilisation�of�the

error-less�codewords.�Development�of�high-order�model�reduction�formulas�is�needed�to�go

beyond�this�first-order�idea�and�reach�the�accuracies�enabling�scalable�quantum�information

processing.

tems with particularly concentratederror channels, thus in creating the con-ditions of the previous paragraph.

This brings us to the more mathematicalchallenges. The design of engineeredreservoirs, such as the singling out of a“quantum machine” from its environ-ment, is based on approximations whereweak couplings are neglected withrespect to dominant effects. Theseapproximations enable tractable sys-tems engineering guidelines to bederived — rather than having to treat asingle big quantum bulk, whose proper-ties are one target of the quantum com-puters themselves. Current designs arebased on setting up systems with firstdominant local effects. This has to be

improved and scaled up to reach therequirements 1. and 2., since every con-trol scheme can only be as precise as itsunderlying model. We are thereforelaunching a concrete effort towardshigh-precision model reduction for-mulas, together with the supercon-ducting circuit experimentalists to iden-tify typical needs, but with a generalscope in mind. Preliminary workalready shows the power of higher-order formulas in identifying designopportunities (Figure 1), paving theway to improved “engineered” hard-ware for robust quantum operation.

Link:

[L1]https://www.inria.fr/en/teams/quantic

References:

[1] M. Mirrahimi, Z.Leghtas et al:“Dynamically protected cat-qubits:a new paradigm for universal quan-tum computation”, New Journal ofPhysics, 2014.

[2] N. Ofek et al.: “Extending the life-time of a quantum bit with errorcorrection in superconducting cir-cuits”, Nature, 2016.

[3] J. Cohen: “Autonomous quantumerror correction with superconductingqubits”, PhD thesis, ENS Paris, 2017.

Please contact:

Alain Sarlette, Pierre Rouchon, [email protected],[email protected]+ 33 1 80 49 43 59


Quantum systems with many degrees offreedom, such as those composed ofmany interacting subsystems or parti-cles, are notoriously hard to simulate onclassical computers. This is because theHilbert space for the quantum states ofthe composite system is exponentiallylarge in the system size, while quantumcorrelations, or entanglement, betweenthe constituent subsystems often pre-clude factorisation of the problem intosmaller parts. This has led RichardFeynman to suggest, back in 1981, tosimulate quantum physics with quantumcomputers, or universal quantum simu-lators [1] composed of many quantumtwo-level systems, like spin-1/2 parti-cles arranged in a lattice, with appropri-ately controlled couplings betweenthem. In principle, any many-bodyHamiltonian dynamics can be decom-posed into small time-steps and finite-range interactions between the qubits.This idea has then developed into a gen-eral purpose quantum computer which issuitable not only for digital simulationsof physical systems, but also forquantum computations. Quantum algo-

rithms for certain mathematical tasks,such as search of unstructured database(Grover) or integer factorisation (Shor),are polynomially or exponentially moreefficient than the best known classicalalgorithms for the same tasks.

Cold atoms trapped in optical lattices orarrays of microtraps represent a scalablearchitecture to realise quantum com-puters as well as analogue and digitalquantum simulations of many-bodydynamics of various spin lattice models.Neutral atoms in the lower electronicstates interact very weakly with eachother, but when excited to the Rydbergstates, their interaction can be verystrong over large distances. Rydbergstates are highly excited states with theatomic electron placed on an orbit thatis far from the ionic core. Due to weakbinding of the electron to the ion, theatoms in the Rydberg states are easilypolarisable. The resulting long-rangeinteractions between the Rydberg atomsmakes them uniquely suited for real-ising strongly-interacting many-bodysystems and for implementing various

quantum information processing tasks[2].

The interatomic interactions are con-trollable by lasers that can excite andde-excite the atoms from the non-inter-acting ground state to the strongly inter-acting Rydberg state on demand. Thishas led to proposals to implementquantum logic gates using the switch-able interactions, or interaction-inducedexcitation blockade, between the atomsrepresenting qubits. There have beenseveral experimental demonstrations ofthe Rydberg quantum gates, but thefidelity of operations has so far beenbelow the threshold value ~0.9999 for ascalable fault-tolerant quantum compu-tation. We study the performance ofquantum algorithms under realisticexperimental conditions involvingnoise and imperfections (see Figure 1).We devise novel schemes for high-fidelity quantum logic gates usingblockade and resonant exchange inter-actions between the Rydberg stateatoms. This work is being done in col-laboration with the theory group of

Quantum Gates and Simulations

with Strongly Interacting Rydberg Atoms

by David Petrosyan (IESL-FORTH)

In order to develop functional devices for quantum computing and analogue and digital quantum

simulations, we need controlable interactions between the physical systems representing

quantum bits – qubits. We explore strong, long-range interactions between atoms excited to high-

lying Rydberg states to implement quantum logic gates and algorithms and to realise quantum

simulators of various spin-lattice models to study few- and many-body quantum dynamics.

Klaus Mølmer at the Aarhus Universityand the experimental group of MarkSaffman at the University ofWisconsin—Madison, USA.

Arrays of trapped atoms excited ondemand to the Rydberg states by laserscan realise various spin-lattice modelsthat can serve for both digital and ana-

logue quantum simulations. In additionto switchable interactions, relaxationsand energy dissipation can be intro-duced in this system in a controlledway. We study the dynamics of few- andmany-body quantum systems usingsuch Rydberg quantum simulators. Asan example, in Figure 2 we show simu-lations of quasi-crystals of Rydbergexcitation as observed in the experi-ments with laser driven atoms in opticallattices [3]. We collaborate with thetheory group of Michael Fleischhauer atthe University of Kaiserslautern,Germany, and part of this work is beingdone within the EU H2020 FETProactive project RySQ (RydbergQuantum Simulations) which involvesmany leading theory and experimentalgroups in Europe.

Finally, strong transitions between theRydberg states of atoms in themicrowave frequency range enabletheir efficient coupling to electrical cir-cuits involving superconducting qubitsand resonators. Superconducting qubitscan realise fast quantum gates but areless suitable for storage of quantuminformation. Coupling atoms to solid-state systems permits the realisation ofhybrid quantum systems composed ofdifferent components with complemen-tary functionalities including quantuminformation processing, storage andconversion to optical photons for long-distance quantum communication. Thisis another direction of our research onquantum computation and simulationswith Rydberg atoms.

Links:

http://www.quantum-technology.gr/http://qurope.eu/projects/rysq/

References:

[1] S. Lloyd: “Universal QuantumSimulators”, Science, Vol. 273(5278), pp. 1073-1078, 1996.

[2] M. Saffman, T. G. Walker, and K.Mølmer, “Quantum informationwith Rydberg atoms”, Rev. Mod.Phys. Vol. 82, pp. 2313-2363, 2010.

[3] P. Schauß et al., “Crystallization inIsing quantum magnets”, ScienceVol. 347 (6229), pp. 1455-1458,2015.

Please contact:

David PetrosyanIESL-FORTH, Greece +30-2810 39 [email protected]



Figure�2:�Realising�interacting�spin�lattice�models�with�laser�driven�atoms.��Upper�panel

illustrates�an�optical�lattice�potential�for�ground�state�atoms�and�laser�coupling�to�the�strongly

interacting�Rydberg�state.�Lower�panel�shows�spatial�configuration�of�six�Rydberg�excitations,

with�the�axial�P(ρ)�and�angular�P(φ)�probability�distributions,�in�a�2D�disk�shaped�lattice�of�400

atoms�driven�by�a�resonant�field.

Figure�1:�Grover�search�algorithm�with�Rydberg�blockaded�atoms.�Left:�Level�scheme�of�the

register�atoms�interacting�with�a�microwave�field�on�the�qubit�transition�and�with�a�resonant

laser�field�on�the�transition�to�the�Rydberg�state.�Atoms�in�Rydberg�states�interact�with�each

other�via�a�strong,�long-range�potential�Vaa which�suppresses�Rydberg�excitation�of�all�but�one

atom�at�a�time.�Right:�Probabilities�of�measuring�correct�outcomes�of�the�Grover�search�versus

number�of�iterations,�for�N=2,3,4�digit�quantum�register,�without�(top)�and�with�(bottom)�decay

of�the�Rydberg�state.�


The goal of quantum control is to designexternal fields (the controls) whoseaction induces a prescribed transitionbetween two states of a quantum system.

A widely used technique is based on theprinciple of adiabatic evolution: If theinitial condition is an eigenstate of thecontrolled Hamiltonian (i.e., theHamiltonian including the external field)and the variation of the external fields isslow, then the state of the system staysclose to the instantaneous eigenstate(namely, the eigenstate of the controlledHamiltonian with frozen time).

Adiabatic evolution works as describedwhen the energy levels of the controlledHamiltonian satisfy a gap condition,which means that they stay away from

one from another and in particular theydo not intersect. The theory can beextended to systems presenting eigen-value intersections. When the intersec-tion of two eigenvalues is transversaland the initial condition corresponds tothe lower eigenvalue then, after theintersection, the trajectory adiabaticallyfollows the eigenstate corresponding tothe upper eigenvalue.

One is typically interested in adiabaticpaths for which the controls start andend at zero. Such controls may be usedto induce nontrivial transitions only ifthere exist paths in the space of controlsthat are not passing back and forththrough the same intersections. This iswhy such an adiabatic setup makessense for control purposes only when

the controls have more than one degreeof freedom.

In the general case one looks at the sin-gular set, i.e., the set of all control valuesfor which the Hamiltonian has degen-erate eigenvalues. The applicability ofthe adiabatic strategy relies on the factthat the singular set does not disconnectthe set of available control parameters.Such a disconnection is very rare, how-ever, in the sense that for a genericsystem the singular set has codimension3 for complex Hamiltonians and codi-mension 2 for real ones.

Figure 1 shows the eigenvalues of a 3-level quantum system driven by twocontrols in the well-known STIRAPconfiguration. In this case there are twoeigenvalue intersections: one betweenthe first and the second levels (on theaxis where the first control is zero) andone between the second and the thirdlevels (where the second control van-ishes). Notice that the resulting adia-batic strategy follows the famouscounter-intuitive pattern, meaning thatone first activates the control associatedwith the transition between the secondand the third energy levels and then theone associated with the transitionbetween first and the second ones, withan overlap between the two pulses [1].

The technique explained above pro-vides a simple strategy to induce a tran-sition between any two eigenstates ofthe free Hamiltonian, under the assump-tion that all energy levels intersect.

In collaboration with FrancescaChittaro (Toulon University) and PaoloMason (CNRS, CentraleSupelec) wehave developed a technique to createsuperpositions between eigenstatesusing broken adiabatic paths [2].

The principle of our approach is that ifan adiabatic path reaches a point in the

Control of Quantum Systems

by broken Adiabatic Paths

by Nicolas Augier (École polytechnique), Ugo Boscain (CNRS) and Mario Sigalotti (Inria)

The dynamics of a quantum mechanical system is described by a mathematical object called

Hamiltonian. The possible results of an energy measure are known as the “eigenvalues” (or energy

levels) of the Hamiltonian. After an energy measure, the system collapses into a particular state

called the eigenstate corresponding to the measured energy. Adding corners to adiabatic paths can

be used to generate superpositions of eigenstates with simple and regular control laws.

Figure�1:�The�eigenvalues�of�a�3-level�quantum�system�driven�by�2�controls�in�the�STIRAP

configuration.

singular set and makes there a corner,then the state splits partially on thelower energy level and partially on theupper one. Modulating the entry andexit direction, hence the angle betweenthe two, one can arbitrarily select theoccupation of the two energy levels.Such an idea makes it possible, startingfrom an eigenstate, to reach any super-position of eigenstates whose corre-sponding eigenvalues intersect [3].

Figure 2, for a 3-level system, shows abroken adiabatic path inducing a transi-

tion from the first energy level to a super-position of the first and the third energylevels. Given the desired populationlevels at the final time, it is possible tocompute the angle between the entry andexit direction at the corner in order toinduce a transition to a state having suchpopulation levels. In the picture, weshow for instance a control inducing atransition from an eigenstate correspon-ding to the first eigenvalue to a superpo-sition with equal weights between thefirst and the third eigenstates, with nopopulation on the second energy level.

The project is supported by the ERCstarting grant GECOMETHODS (2010-2016) and by the French ANR projectsGCM (2009-2013). Extensions and algo-rithmic refinements are under study inthe framework of the Inria team CAGEthanks to the support of the French ANRproject QUACO (2017-2021).

References:

[1] C.E. Carroll and F.T. Hioe:“Analytic solutions for three-statesystems with overlapping pulses”,Phys. Rev. A, vol. 42, pp. 1522-1531, 1990.

[2] U. Boscain et al.: “Adiabaticcontrol of the Schrödinger equationvia conical intersections of theeigenvalues”, IEEE Trans. Autom.Control, vol. 57, pp. 1970-1983,Aug. 2012.

[3] U. Boscain et al.: “Approximatecontrollability, exact controllability,and conical eigenvalue intersectionsfor quantum mechanical systems”,Commun. Math. Phys, vol. 333, pp.1225-1239, Feb. 2015.

Please contact:

Nicolas Augier, CMAP, Écolepolytechnique, [email protected]

Ugo Boscain, CNRS, LJLL, UniversitéPierre et Marie Curie, [email protected]

Mario SigalottiInria, LJLL, Université Pierre et MarieCurie, [email protected]



Figure�2:�A�broken�adiabatic�path�inducing�a�transition�from�the�first�energy�level�to�a

superposition�of�the�first�and�the�third�energy�levels.

the Case for Quantum Software

by Harry Buhrman (CWI and QuSoft) and Floor van de Pavert (QuSoft)

Researchers and industry specialists across Europe have launched a Quantum Software

Manifesto. With the Manifesto, the group aims to increase awareness of and support for

quantum software research.

Quantum computers, once just thedream of science fiction writers, are rap-idly becoming a reality. Already, the firstsmall quantum hardware devices arebeing put through their paces, withresearchers probing for evidence thatthey really do work in a fundamentallydifferent way, unlocking solutions toproblems classical computers couldnever solve.

Several leading scientists and decisionmakers, recognising the imminent prac-tical impact of quantum technologies,came together in 2016 to write theQuantum Manifesto [L1] (not to be con-fused with the Quantum SoftwareManifesto), calling for an ambitiousEuropean quantum technology initiativethat would place Europe at the heart ofthese new developments. After more

than 3,400 endorsements from people inscientific, industrial and governmentalorganisations, the European Commissionlaunched the Flagship Initiative onQuantum Technologies.

As miraculous as the new quantum hard-ware may be, though, it can never reachits full potential without great quantumsoftware; this new paradigm will


demand new approaches, algorithms andprotocols. With this in mind, and spurredby feeling that the Flagship Initiativerisks under-representing the crucial roleof software and theory, we have devel-oped the Quantum Software Manifesto,which focuses on the status, outlook andspecific challenges facing the quantumsoftware field.

It is important to remember, first of all,that quantum computers are no silverbullet: where some problems will seedramatic speedups, others see none atall. Identifying potential applicationsand designing new quantum algorithmsbased on new principles will be critical,as will finding ways to achieve a usefulquantum advantage on the small, noisydevices that are likely to be available inthe short to medium term. We need bothfoundational, theoretical breakthroughsand practical work on optimising algo-rithms for real quantum architectures.That, in turn, will demand ever-closercollaboration between software andhardware developers, and between aca-demic and industrial partners.

In particular, developing reliablequantum computers is exceedingly diffi-cult because their basic building blocks,called qubits, are so fragile. Even thetiniest environmental perturbation candestroy their delicate superpositionstates. Software techniques for quantumerror correction and fault-tolerant com-putation could greatly ease the task ofhardware design, helping to pave theway for large, stable quantum systems.New verification and testing protocolsbased on new theoretical ideas will alsobe essential to both guarantee than suchdevices are functioning correctly andguide their design.

Important practical applications includesimulating physical and chemical sys-tems, approximate optimisation andmachine learning. Classical computersfind it notoriously difficult to simulatequantum systems; indeed, this currentlytakes up about 20% of supercomputertime. In contrast, little could be morenatural for a quantum computer, and thiswill be vital in fields such as quantumchemistry, materials science and high-energy physics. Many exciting experi-ments are already bringing this ideacloser to reality.

Machine learning is another hot topic inthe modern world, and a flurry of new

developments in this area have beenkicked off by the discovery of an expo-nential quantum speedup for solvingparticular types of linear equations. Thishas led to new algorithms for core prob-lems such as data fitting, support vectormachines and classification. Indeed, anew quantum recommender can alreadyhelp users of, say Netflix or Amazon, tofind good options exponentially faster.But the real work has still has to begin infinding applications where these tech-niques can be exploited.

Quantum algorithms have the potentialto both break current cryptography (viaShor’s algorithm) and provide new andfundamentally more secure cryptosys-tems, in principle even allowing users torun programs on untrusted systemswhile keeping their data secret. Practicaland commercial technologies havealready been developed for quantum keydistribution and random number genera-tion, but much work remains to fullydevelop the potential of quantum cryp-tography, particularly for large-scalequantum networks.

Quantum networking has already beendemonstrated, with entangled statesbeing successfully distributed fromsatellites to ground stations. As well asallowing truly secure communication, aglobal quantum internet would enabledistributed quantum algorithms toexploit the power of quantum teleporta-tion and error correction to solve distrib-uted tasks that require quantumresources or for which they can makethe communication more efficient..

If any of these developments can cometo pass, however, we will need moregood quantum programmers. Becauseworking with quantum computers is sofundamentally different from classicalprogramming, the pool of people withthe necessary knowledge is currentlysmall and new education programs areurgently needed, in both academia andindustry. Although some initial stepshave been taken, a proposed curriculumis still in its infancy.

We stand at the dawn of the quantumera. It may be imminent, or may take alittle while longer to materialise, and webelieve it is investment in quantum soft-ware and programmers that will makethe difference. This will help us to buildpractical hardware and develop life-changing applications, but none that willbe possible without an integratedapproach to quantum hardware and soft-ware development by industry and aca-demia.

The Quantum Software Manifesto canbe downloaded via [L2].

Link:

[L1] https://kwz.me/hBj[L2] https://kwz.me/hBq

Please contact:

Harry Buhrman, QuSoft and CWI, The NetherlandsFloor van de Pavert QuSoft, The Netherlands+31 (0)20 592 [email protected]

The�Quantum�Software

Manifesto�tcalls�for

increased�awareness�of�the

importance�and�urgency�of

quantum�software

research.

Eu

rop

ea

n R

ese

arc

h a

nd

In

no

va

tio

n

Research and Innovation

Computers that Negotiate

on our behalf

by Tim Baarslag (CWI)

Computers that negotiate on behalf of humans hold

great promise for the future and will even become

indispensable in emerging application domains such as

the smart grid, autonomous driving, and the Internet of

Things. An important obstacle is that in many real-life

settings, it is impossible to elicit all information

necessary to be sensitive to the individual needs and

requirements of users. This makes it a lot more

challenging for the computer to decide on the right

negotiation strategy; however, new methods are being

created at CWI that make considerable progress towards

solving this problem.

Imagine a system that helps a group of friends decide on aholiday destination based on their individual preferencesabout cultural activities, costs, and flight duration. After that,the system contacts a number of travel operators to negotiatethe best hotel and airline deal and proposes a joint holidayschedule. This is what researchers believe we can one dayexpect from the research field of automated negotiation: adomain of mathematics and computer science that designsalgorithms that can negotiate with people and among com-puterised systems. As part of the NWO InnovationalResearch Incentives Scheme Veni and the EU project Grid-Friends, CWI is developing technology that can help com-puters get the best deal for users, even when their preferencesare not fully specified.

The field of automated negotiation is fueled by a number ofbenefits that computerised negotiation can offer, includingbetter (win-win) and faster deals, and reduction costs, stressand cognitive effort on the part of users [1]. Autonomousnegotiation technology might even play an indispensablerole in real-world applications where the human scale issimply too slow and expensive. For instance, with the world-wide deployment of the smart electrical grid and the must forrenewable energy sources, flexible devices in our householdwill soon (re-)negotiate complex energy contracts automati-cally. Another example is the rise of the Internet of Things(IoT), which will introduce countless smart, interconnecteddevices that autonomously negotiate the usage of sensitivedata and make trade-offs between privacy concerns, price,and convenience. In such settings, the agent can help repre-sent users in complex and constantly ongoing negotiations inan automated manner.

However, one of the key challenges in designing a successfulautomated negotiator is that in real-life settings, only limitedinformation is available about the user and other parties.Users are often unwilling or unable to fully specify theirpreferences to a negotiation system; as a consequence, auto-mated negotiators are required to strike deals with very lim-ited available user information. Recently, we investigatedsuch negotiations in two different domains: privacy negotia-tions and smart grid trading.



Together with The University of Southampton and TheMassachusetts Institute of Technology, we are working onnew interaction mechanisms for achieving mutually benefi-cial agreements, i.e., negotiation, as a more flexible interac-tion paradigm for meaningful consent towards data sharingand permission management [3]. The aim is to address theinadequacy of current interaction mechanisms for handlingdata requests and obtaining consent, such as cookie noticesand permission pop-ups, which fail to align consumerchoices with their privacy preferences. The hope is to pro-vide users with a more granular and iterative permissionmodel than current take-it-or-leave it approaches. The maincatalyst for improvement is a new querying model that canelicit preference information at the right time based on infor-mation theoretical models from search theory. The firstresults from our lab study show that users are able andwilling to share significantly more of their personal datawhen offered the benefits of negotiation, while maintainingthe same level of satisfaction with their sharing decisions.Users adopt strategies in which they explore around theirdesired permission set for a more nuanced negotiation thatbetter aligns with their privacy preferences.

As a second line of research, we explore automated negotia-tions within the smart electrical grid as part of the Grid-Friends project, which is coordinated by CWI in cooperationwith Fraunhofer-ITWM. The Grid-Friends team is currentlydeveloping efficient algorithms that can be used for user-

adaptable energy management systems within a smart gridcooperative of homeowners. We developed an optimal andtractable decision model based on adaptive utility elicitation[2] that can find the point of diminishing returns forimproving the model of user preferences. Our frameworkprovides an extensible basis for interactive negotiationagents and is scheduled to be put in practice in 2018 within ahousehold community in Amsterdam.

Collaborative work is currently being undertaken to extendthis research further. We are organising a yearly automatednegotiation competition (ANAC) [L1, L2] where uncertainpreferences will act as a novel challenge for the negotiationresearch community. Other future work will include person-alised assistants (including for travel), autonomous driving,and making meaningful and dynamic consent workable atthe Internet of Things scale.

This work is part of the Veni research programme withproject number 639.021.751, which is financed by theNetherlands Organisation for Scientific Research (NWO).This work is supported by a grant from the ESPRC forMeaningful Consent in the Digital Economy project(EP/K039989/1). The research has received funding throughthe ERA-Net Smart Grids Plus project Grid-Friends, withsupport from the European Union’s Horizon 2020 researchand innovation programme.

References:

[1] T. Baarslag, et al.: “Computers That Negotiate on OurBehalf: Major Challenges for Self-sufficient, Self-directed, and Interdependent Negotiating Agents”,AAMAS 2017 Workshops Visionary Papers, LectureNotes in Computer Science. Springer International Pub-lishing, Cham, 2017.

[2] T. Baarslag and M. Kaisers: “The value of informationin automated negotiation: A decision model for elicitinguser preferences”, in Proc. of the 16th Conf. onAutonomous Agents and MultiAgent Systems,AAMAS’17, p. 391-400, Richland, SC, 2017. Interna-tional Foundation for Autonomous Agents and Multia-gent Systems.http://dl.acm.org/citation.cfm?id=3091125.3091185

[3] T. Baarslag, Alper T. Alan, Richard C. Gomer, IlariaLiccardi, Helia Marreiros, Enrico H. Gerding, and M.C.Schraefel.et al.: “Negotiation as an interaction mecha-nism for deciding app permissions”, in Proc. of the2016 CHI Conference: Extended Abstracts on HumanFactors in Computing Systems, CHI EA’16, p. 2012-2019, New York, NY, USA, 2016. ACM.http://doi.acm.org/10.1145/2851581.2892340

Please contact:

Tim Baarslag, CWI, The Netherlands+31(0)20 592 4019, [email protected]

37

Figure�1:�What�if�we�could�negotiate�our�app�permissions?


faster text Similarity

Using a Linear-Complexity

Relaxed Word Mover’s

distance

by Kubilay Atasu, Vasileios Vasileiadis, Michail Vlachos(IBM Research – Zurich)

A significant portion of today’s data exists in a textual

format: web pages, news articles, financial reports,

documents, spreadsheets, etc. Searching across this

collected text knowledge requires two essential

components: a) A measure to quantify what is

considered ‘similar’, to discover documents relevant to

the users’ queries, b) A method for executing in real-time

the similarity measure across millions of documents.

Researchers in the Information Retrieval (IR) communityhave proposed various document similarity metrics, either atthe word level (cosine similarity, Jaccard similarity) or at thecharacter level (e.g., Levenshtein distance). A major short-coming of traditional IR approaches is that they identify doc-uments with similar words, so they fail in cases when similarcontent is expressed in a different wording. Strategies to mit-igate these shortcoming capitalize on synonyms or conceptontologies, but maintaining those structures is a non-trivialtask.

Artificial Intelligence and Deep Learning have heralded anew era in document similarity by capitalizing on vastamounts of data to resolve issues related to text synonymyand polysemy. A popular approach, called ‘word embed-dings’, is given in [1], which maps words to a new space inwhich semantically similar words reside in proximity to eachother. To learn the embedding (i.e., location) of the words inthe new space, those techniques train a neural network usingmassive amounts of publicly available data, such asWikipedia. Word embeddings have resolved many of theproblems of synonymy.

In the new embedding space, each word is represented as ahigh-dimensional vector. To discover documents similar to auser’s multi-word query, one can use a measure called the‘Word-Mover’s Distance’, or WMD [2], which essentiallytries to find the minimum cost to transform one text toanother in the embedding space. An example of this is shown

in Figure 1, which can effectively map the sentence “TheQueen to tour Canada” to the sentence “Royal visit toHalifax”, even though (after “stopword” removal) they haveno words in common. WMD has been shown to work veryeffectively in practice, outperforming both in precision andin recall many traditional IR similarity measures. However,WMD is costly to compute because it solves a minimum costflow problem, which requires cubic time in the averagenumber of words in a document. The authors in [2] have pro-posed a lower-bound to the WMD called Relaxed WDM, orRWMD, which in practice, retrieves documents very similarto WMD, but at a much lower cost. However, RWMD stillexhibits quadratic execution time complexity and can provecostly when searching across millions of documents.

For searches involving millions of documents, the RWMDexecution is inefficient because it may require repetitivecomputation of the distance across the same pairs of words.Pre-computing the distances across all words in the vocabu-lary is a possibility, but it is wasteful regarding storage space.To mitigate these shortcomings, we have proposed a linear-complexity RWMD that avoids wasteful and repetitive com-putations. Given a database of documents, where the docu-ments are stored in a bag-of-words representation, the linear-complexity RWMD computes the distance between a querydocument and all the database documents in two phases:• In the first phase, for each word in the vocabulary, the

Euclidean distance to the closest word in the query docu-ment is computed based on the vector representations ofthe words. This phase involves a dense matrix-matrix mul-tiplication followed by a column- or row-wise reduction tofind the minimum distances. The result is a dense vector Z.

• In the second phase, the collection of database documentsis treated as a sparse matrix, which is then multiplied withthe dense vector Z. This phase produces the RWMD dis-tances between the query document and all the databasedocuments.

Both phases map very well onto the linear algebra primitivessupported by modern GPU (Graphics Processing Unit)devices. In addition, the overall computation can be scaledout by distributing the query documents or the reference doc-uments across several GPUs for parallel execution. Figure 2depicts the overall approach, which enables the linear-com-plexity RWMD to achieve high speeds.

The result is that, when computing the similarity of one doc-ument with h words against n documents each having onaverage h words using a vector space of dimension m, thecomplexity of the brute-force RWMD is O(nh2m), whereas


Figure�1:�(Left)�Mapping�of�word

to�a�high-dimensional�space�using

word�embeddings.�(Right)�An

illustration�of�the�Word�Mover’s

Distance.


under our methodology the complexity is O(nhm). The inter-ested reader can find additional technical details about thelinear complexity RWMD implementation in [3].

To showcase the performance of the new methodology, wehave used very large datasets of news documents. One docu-ment is posed as the query and the search retrieves the k-most-similar documents in the database. Such a scenario canbe used either for performing duplicate detection (when thedistance is below a threshold), or for achieving clustering ofsimilar news events. We conducted our experiments on acluster of 16 NVIDIA Tesla P100 GPUs on two datasets. Set1 comprises 1 million documents, with an average number of108 words per document. Set 2 comprises 2.8 million words,with an average number of 28 words per document. Thecomparison of the runtime performance between WMD,RWMD, and our solution is given in Figure 3, which showsthat the proposed linear-complexity RWMD can be morethan 70 times faster than the original RWMD and more than16,000 times faster than the Word Mover’s Distance.

The results suggest that we could use the high-qualitymatches of the RWMD to query – in sub-second time – atleast 100 million documents using only a modest computa-tional infrastructure.

References:

[1] T. Mikolov, K. Chen, G. Corrado, and J. Dean: “Effi-cient estimation of word representations in vectorspace”, CoRR, abs/1301.3781, 2013.

[2] M. J. Kusner, et al.: “From word embeddings to docu-ment distances, ICML 2015: 957–966.

[3] K. Atasu, et al.: “Linear-Complexity Relaxed WordMover’s Distance with GPU Acceleration”, IEEE BigData 2017.

Please contact:

Kubilay Atasu, IBM Research – Zurich, [email protected]

39

Figure�2:�

Overview�of�our�approach.

!

•!

. . .

Query documents (Q)

Phase 1

Clu

ster

of G

PUs

!"#$%&'($)"#*$+,$-(%$."'&/01&#23$'"450-%$-(%$*+6-&,'%$-"$-(%$'1"6%6-$)"#*$+,$",% 70%#2 *"'8$9-"#%$-(%$#%601-6$+,$&$*%,6%$.%'-"#$:8

Phase 2

Dense vector Z

95&#6%;4&-#+<$*%,6%;.%'-"#$401-+51+'&-+",$/%-)%%,$= &,*$:$ -"$'"450-%$-(%$*+6-&,'%6$/%-)%%,$-(%$70%#2$&,*$-(%$*&-&/&6%$*"'68

Database documents (D)

Distribute Qacross GPUs

!

!

$ $ $ $$ $

$ $$ $ $$ $ $

$$ $

$ $ $$

$$

$ $

$ $$

!

!

$ $ $ $$ $

$ $$ $ $$ $ $

$$ $

$ $ $$

$$

$ $

$ $$

!

!

$ $ $ $$ $

$ $$ $ $$ $ $

$$ $

$ $ $$

$$

$ $

$ $$

!

!

$ $ $ $$ $

$ $$ $ $$ $ $

$$ $

$ $ $$

$$

$ $

$ $$

!

!

$ $ $ $$ $

$ $$ $ $$ $ $

$$ $

$ $ $$

$$

$ $

$ $$

!

!

$ $ $ $$ $

$ $$ $ $$ $ $

$$ $

$ $ $$

$$

$ $

$ $$

1esaPh

,+*#")'(&%#"!32#&10/'&."%(-

Dat

!

!

$ $ $ $$ $

$ $$ $ $$ $ $

$$ $

$ $ $$

$$

$ $

$ $$

saPh

9 6%#&5 ;6%,%* ;.%

aseabat )D(snteumdoc

!

!

$ $ $ $$ $

$ $$ $ $$ $ $

$$ $

$ $ $$

$$

$ $

$ $$

2e

<$+#-&4#"'-.%

!

!

$ $ $ $$ $

$ $$ $ $$ $ $

$$ $

$ $ $$

$$

$ $

$ $$

!

!

$ $ $ $$ $

$ $$ $ $$ $ $

$$ $

$ $ $$

$$

$ $

$ $$

Q yrue nteumdoc

!

!

$ $ $ $$ $

$ $$ $ $$ $ $

$$ $

$ $ $$

$$

$ $

$ $$

snt )(Q

Us

PGfo

2%(-%-054'"%(-"-'%,&6-+*,+*#")6-6%"'1

",% 270%# '8"*6$-160%#%(-%#"-9

6%,%*&,+8:#"'-.%

!

!

$ $ $ $$ $

$ $$ $ $$ $ $

$$ $

$ $ $$

$$

$ $

$ $$

seneDZroctve

'&+15+-104%%,$/%-)

:$ 4'""--(%$ &6-+*

%%,$/%-),&2$#%076%&/&-&*

!

!

$ $ $ $$ $

$ $$ $ $$ $ $

$$ $

$ $ $$

$$

$ $

$ $$

,"+-'&%%,$= &,*$

%-0546$'%,&

%%,$-(%$%(-*,'68"*6%

!

!

$ $ $ $$ $

$ $$ $ $$ $ $

$$ $

$ $ $$

$$

$ $

$ $$

birstiDa ss ocr

!

!

$ $ $ $$ $

$ $$ $ $$ $ $

$$ $

$ $ $$

$$

$ $

$ $$

retsuC

letu Q

sUPG

!

!

$ $ $ $$ $

$ $$ $ $$ $ $

$$ $

$ $ $$

$$

$ $

$ $$

.. .

!

!

$ $ $ $$ $

$ $$ $ $$ $ $

$$ $

$ $ $$

$$

$ $

$ $$

!

!

$ $ $ $$ $

$ $$ $ $$ $ $

$$ $

$ $ $$

$$

$ $

$ $$

!

!

$ $ $ $$ $

$ $$ $ $$ $ $

$$ $

$ $ $$

$$

$ $

$ $$

!

!

$ $ $ $$ $

$ $$ $ $$ $ $

$$ $

$ $ $$

$$

$ $

$ $$

!

!

$ $ $ $$ $

$ $$ $ $$ $ $

$$ $

$ $ $$

$$

$ $

$ $$

!

!

$ $ $ $$ $

$ $$ $ $$ $ $

$$ $

$ $ $$

$$

$ $

$ $$

!

!

$ $ $ $$ $

$ $$ $ $$ $ $

$$ $

$ $ $$

$$

$ $

$ $$

!

!

$ $ $ $$ $

$ $$ $ $$ $ $

$$ $

$ $ $$

$$

$ $

$ $$

!

!

$ $ $ $$ $

$ $$ $ $$ $ $

$$ $

$ $ $$

$$

$ $

$ $$

!

!

$ $ $ $$ $

$ $$ $ $$ $ $

$$ $

$ $ $$

$$

$ $

$ $$

!

!

$ $ $ $$ $

$ $$ $ $$ $ $

$$ $

$ $ $$

$$

$ $

$ $$

!

!

$ $ $ $$ $

$ $$ $ $$ $ $

$$ $

$ $ $$

$$

$ $

$ $$

[

!"#$% !"#$&

'

'('%

'(%

%

%'

%''

)*+#,-"$./-012,3/+

456

)456

7.8)456

9,-

"$:3";3<

&=>?

@%? &&?

A@?

Figure�3:�Time�to�compute�the�k-most-

similar�documents�using�16�NVIDIA�Tesla

P100�GPUs.


the CAPtCHA Samples

Website

by Alessia Amelio (University of Calabria), Darko Brodić,Sanja Petrovska (University of Belgrade), RadmilaJanković (Serbian Academy of Sciences and Arts)

“CAPTCHA Samples” is a new website for testing

different types of CAPTCHA specifically designed for

research and study purposes.

In recent decades, the CAPTCHA test has received muchattention owing to the increasing security problems affectingthe web, for which risk prevention plays a key role.CAPTCHA is an acronym of “Completely Automated PublicTuring Test to tell Computers and Humans Apart”. This is atest program which an internet user is required to solve inorder to prove that he or she is human and thus gain access toa given website. In fact, the CAPTCHA test is designed to beeasily solved by a human and difficult to solve for an auto-matic program which tries to obtain unauthorized access tothe website [1]. Recently, different types of CAPTCHA haveflourished in the literature, with more sophisticated securitymechanisms that guarantee efficacy in risk prevention, andmore user-friendly interfaces [2], [3]. In order for researchersto effectively build on the recent work in this area, an analysisof the most recently introduced CAPTCHA tests is needed.

Accordingly, we present CAPTCHA Samples, a new websitecollecting different CAPTCHA tests for research andanalysis purposes. CAPTCHA Samples, available at [L1](see Figure 1), was born as a joint project involving theTechnical Faculty in Bor, University of Belgrade, Serbia, andDIMES University of Calabria, Italy.

Currently, the website reports different samples of image-based CAPTCHA tests, including tests based on facialexpressions (animated characters, face of an old woman, sur-prised face, and worried face). The aim of these CAPTCHAtests is to recognise the correct image from a list of differentimages, according to the question asked by the test. From thehomepage, it is possible to select the CAPTCHA test ofinterest and access to the corresponding webpage. TheCAPTCHA test can be solved in order to obtain useful infor-mation about the usability of the test, including: (i) comple-tion time, which is the solution time to the CAPTCHA, and(ii) number of tries to provide the correct solution to theCAPTCHA.

If the CAPTCHA test is not correctly solved, then the failure isnotified by a text field of “wrong answer” on the screen, andthe user is asked to try again. Otherwise, if the CAPTCHA testis correctly solved, a message in a text field on the screen isvisualised, reporting the completion time and the number of

tries to find the correct solution. Figure 2 shows an example ofa successfully solved “surprised face” CAPTCHA test, whichwas solved in 2.48 seconds and one attempt.

Every test is completely anonymous because any personalinformation is registered about the internet user whoaccesses the CAPTCHA. Also, the main advantage of thewebsite is that the CAPTCHA tests can be accessed from dif-ferent devices, including smartphones, laptops and tablets. Inthis way, the difference in terms of CAPTCHA usabilitybetween multiple devices can be easily checked. In thefuture, we are planning to extend the CAPTCHA Sampleswebsite with new functionalities, including the selection ofother CAPTCHA tests of different typology, which will beadded to the website. It will be particularly useful in aca-demic as well as industrial research for anonymously gath-ering data about the usability of CAPTCHA tests. It couldalso be the starting point for designing new CAPTCHA typeswhich are appropriate for specific types of internet users. Formore detailed information about the image-basedCAPTCHA tests included in the CAPTCHA Samples web-site, please refer to [3].

Link:

[L1] http://captchasamples.altervista.org

References:

[1] A.M. Turing: “Computing Machinery and Intelligence”,Mind 59:433-460, 1950.

[2] D. Brodić, S. Petrovska, M. Jevtić, Z. N. Milivojević:“The influence of the CAPTCHA types to its solvingtimes”, MIPRO: 1274-1277, 2016.

[3] D. Brodić, A. Amelio, R. Janković: “Exploring theInfluence of CAPTCHA Types to the Users ResponseTime by Statistical Analysis”, Multimedia Tools andApplications, 1-37, 2017.

Please contact:

Alessia Amelio, University of Calabria, Rende, [email protected]


Figure�2:�Webpage�of�CAPTCHA�Samples�with�a�message�reporting

success�in�solving�the�test,�including�completion�time�and�number�of

attempts.

Figure�1:�CAPTCHA

Samples�website.�


APoPSIS: A Web-based

Platform for the Analysis

of Structured dialogues

by Elisjana Ymeralli, Theodore Patkos and YannisRoussakis (ICS-FORTH)

APOPSIS is a web-based platform that aims to motivate

online users to participate in well-structured dialogues by

raising issues and posting ideas or comments, related to

goal-oriented topics of discussion. The primary goal of the

system is to offer automated opinion analysis features

that help identify useful patterns of relations amongst

participants and their expressed opinions. Our system is

designed to enable more structured and less confusing

argumentative discussions, thus helping sense-makers in

understanding the dynamic flow of the dialogue.

Web-based platform APOPSIS can be used in everydaydeliberation, as well as decision-making discussions, whereusers exchange their viewpoints and argue over a plethora oftopics. The platform offers well-structured dialogues whichcan take place in different phases of discussions. As adebating platform, APOPSIS offers theopportunity for a variety of groups towork and collaborate with each other bysharing their ideas in support of or againstother opinions. Each dialogue is open forusers to defend and justify their state-ments and vote upon them, respectively.The first phase of the dialogue allowsusers to participate by providing solu-tions or statements, in addition to theirjustified agreements or disagreements,while in the second phase, only the mostpopular or well-justified solutions remainand become subject to more detailedevaluation by the participants.

Conversations are presented in a tree-likestyle, where subsequent levels of com-ments respond to the parent comment.The system supports several features thatenrich the system’s usability, such asvoting, semantic annotation, aspect-basedrating and searching functionalities.These features enable APOPSIS to pro-duce and extract useful conclusions thatcan help sense-makers and expert users tounderstand and make decisions on spe-cific problems and issues.

In order to interpret and analyse userreactions, such as comments, votes andratings, the system uses methods from thefields of computational argumentationand machine learning (ML) that enablethe structuring and evaluation of the dif-fering opinions expressed in dialogues.Specifically, an argumentation-based

approach that has its roots in the IBIS-style argumentationmodel [1], but with slightly different semantics, is designedto structure the argumentation process [2] and organise thedifferent conceptual components of the dialogue. In order tobe able to accommodate (store and retrieve) all these compli-cated opinions, a formal semantic web ontology, calledMACE (Multi Aspect Comment Evaluation), was designed,which can semantically represent the content produced andexchanged within the platform for online communities.Moreover, the system uses the s-mDiCE (symmetric multi-Dimensional Comment Evaluation) argumentation frame-work [3], for evaluating online discussions and identifyingthe strongest and most acceptable opinions found in onlinedebates, based on different metrics. The underlying quantita-tive algorithms are generic enough to capture the features ofonline communities on the social web and may benefit sev-eral platforms, from debate portals to decision-making sys-tems.

A key feature of APOPSIS is the support for automatedopinion analysis for analysing user behaviour in online com-munities. Opinion analysis aims at identifying differentgroups of users and useful patterns of relationship betweenparticipants and opinions, in order to help users to makebetter sense of the dialogue and the opinion exchangeprocess. Based on user reactions, the system applies ML

41

Figure�1:�A�part�of�City�Planning�dialogue.

Figure�2:�An�automated�Opinions�Analysis.


algorithms for the clustering of features and the extraction ofassociation rules.

The basic ML algorithms used in this platform are:• Expectation-Maximisation algorithm (EM): Decides the

optimal number of clusters. • K-means algorithm: Identifies different groups of users

and opinions.• Apriori algorithm: Determines interesting and useful rela-

tionships among attributes.

Our methodology implements five different informationneeds emerging from users throughout the sense-makingprocess.• Same profiles with similar opinions: Identifies different

user profiles that share similar opinions on specific posi-tions.

• Sharing similar opinions with specific user: Extracts dif-ferent groups of users whose opinions are closely relatedto specific user(s).

• Same users with similar preferences: Determines groupsof users who share both similar opinions and profile char-acteristics.

• Similar opinions based on different profiles: Identifiessimilar opinions expressed by users with different profilecharacteristics.

• Different user profiles with similar/dissimilar opinions:Finds different groups of user profiles who share similaror dissimilar opinions.

There are several avenues that are worthy of further investi-gation, with an emphasis on improving the system’susability. Thorough evaluations with real users and largedatasets of discussions will be carried out, including expertwalk-through evaluation of the platform and adjustmentsbased on user expert feedback.

Links:

[L1] www.ics.forth.gr/isl/apopsis[L2] http://www.ics.forth.gr/isl/mace/mace.rdfs[L3] http://www.ics.forth.gr/isl/mace/MACE%20Ontol-ogy_Scope_Notes.pdf

References:

[1] J. Conklin, M.L. Begeman: “gIBIS: A hypertext tool forteam design deliberation”, Proc. of the ACM confer-ence on Hypertext, ACM, 1987.

[2] J.Schneider, T.Groza, and A.Passant: “A review of argu-mentation for the social semantic web”, Semantic Web4.2 (2013)

[3] T.Patkos, G.Flouris, and A.Bikakis: “Symmetric Multi-Aspect Evaluation of Comments”, ECAI 2016-22ndEuropean Conference on Artificial Intelligence, 2016.

Please contact:

Elisjana Ymeralli, FORTH, [email protected] (mailto:[email protected])

Can we trust Machine

Learning Results?

Artificial Intelligence

in Safety-Critical decision

Support

by Katharina Holzinger (SBA Research), Klaus Mak,(Austrian Army) Peter Kieseberg (St. Pölten University ofApplied Sciences), Andreas Holzinger (Medical UniversityGraz, Austria)

Machine learning has yielded impressive results over the

last decade, but one important question that remains to

be answered is: How can we explain these processes and

algorithms in order to make the results applicable as

proof in court?

Artificial intelligence (AI) and machine learning (ML) aremaking impressive impacts in a range of areas, such asspeech recognition, recommender systems, and self-drivingcars. Amazingly, recent deep learning algorithms, trained onextremely large data sets, have even exceeded human per-formance in visual tasks, particularly on playing games suchas Atari Space Invaders, or mastering the game of Go [L1].An impressive example from the medical domain is therecent work by Esteva et al. (2017) [1]: they utilised aGoogleNet Inception v3 convolutional neural network(CNN) architecture for the classification of skin lesions, pre-trained their network with approximately 1.3 million images(1,000 object categories), and trained it on nearly 130,000clinical images. The performance was tested against 21board-certified dermatologists on biopsy-proven clinicalimages. The results show that deep learning can achieve aperformance on par with human experts.

One problem with such deep learning models is that they areconsidered to be “black-boxes”, lacking explicit declarativeknowledge representation, hence they have difficulty in gen-erating the required underlying explanatory structures. Thisis limiting the achievement of their full potential, and even ifwe understand the mathematical theories behind the machinemodel, it is still complicated to get insight into the internalworkings. Black box models lack transparency and one ques-tion is becoming increasingly important: “Can we trust theresults?” We argue that this question needs to be rephrasedinto: “Can we explain how and why a result was achieved?”(see Figure 1). A classic example is the question “Whichobjects are similar?” This question is the typical pretext forusing classifiers to classify data objects, e.g. pictures, intodifferent categories (e.g., people or weapons), based on util-ising implicit models derived through training data. Still, aneven more interesting question, both from theoretical andlegal points of view, is “Why are those objects similar?” Thisis especially important in situations when the results of AIare not easy to verify. While verification of single items iseasy enough in many classical problems solved withmachine learning, e.g., detection of weapons in pictures, itcan be a problem in cases where the verification cannot be



done by a human with (close to) 100 percent precision, as incancer detection, for example.

Consequently, there is growing demand for AI, which notonly performs well, but is also transparent, interpretable andtrustworthy. In our recent research, we have been working onmethods and models to reenact the machine decision-makingprocess [2], to reproduce and to comprehend the learning andknowledge extraction processes, because for decision sup-port it is very important to understand the causality oflearned representations. If human intelligence is comple-mented by machine learning, and in some cases even over-ruled, humans must be able to understand, and most of all tobe able to interactively influence the machine decisionprocess. This needs sense making to close the gap betweenhuman thinking and machine “thinking”.

This is especially important when the algorithms are used toextract evidence from data to be used in court. A similar dis-cussion has already been started in the area of digital foren-sics, with novel forensic processes being able to extract smallparts of information from all over the IT-system in questionand then combining them in order to generate evidence –which is currently not usable in court, simply owing to thefact that no judge will rule on evidence gathered by a processno one can control and see through.

Our approach will also address rising legal and privacy con-cerns, e.g., with the new European General Data ProtectionRegulation (GDPR and ISO/IEC 27001) entering into force onMay, 25, 2018. This regulation will make black-boxapproaches difficult to use in business, because they are notable to explain why a decision has been made. In addition, itmust be noted that the “right to be forgotten” [3] established bythe European Court of Justice has been extended to become a“right of erasure”; it will no longer be sufficient to remove aperson’s data from search results when requested to do so, datacontrollers must now erase that data. This is especially prob-lematic when data is used inside the knowledge base of amachine learning algorithm, as changes here might make deci-sions taken and results calculated by the algorithm irrepro-ducible. Thus, in order to understand the impact of changes tosuch results, it is of vital importance to understand the inter-

nals of the intrinsic model built by thesealgorithms. Furthermore, and in spirit withour second research interest in this area,the deletion of data is a highly complicatedprocess in most modern complex environ-ments and gets even more complicatedwhen considering the typical targets ofdata provisioning environments like data-bases that are opposing deletion:• Fast searches: Typically, one main goal

in database design is to provide fastand efficient data retrieval. Thus, theinternal structures of such systemshave been designed in order to speedup the search process by incorporatingparts of the content as search keys,yielding a tree structure that is organ-ised along the distribution of key infor-mation, thus making deletion a prob-lematic issue.

• Fast data manipulation: Like in modern file systems, dataentries that are “deleted” are not actually erased from thedisk with overwriting the respective memory for perform-ance reasons, but only unlinked from the search indicesand marked for overwriting.

• Crash Recovery: Databases must possess mechanisms incase an operation fails (e.g., due to lack of disk space) orthe database crashes in the middle of a data-altering oper-ation (e.g., blackouts) by reverting back to a consistentstate that is throughout all data tables. In order to providethis feature, transaction mechanisms must store the dataalready written to the database in a crash-safe manner,which can be used, for example, in forensic investigationsto uncover deleted information.

• Data Replication: Companies have implemented mirrordata centres that contain replicated versions of the opera-tive database in order to be safe against failures. Deletionfrom such systems is thus especially complicated.

With our research, we will be able to generate mechanismsfor better understanding and control of the internal models ofmachine learning algorithms, allowing us to apply fine-grained changes on one hand, and to better estimate theimpact of changes in knowledge bases on the other hand. Inaddition, our research will yield methods for enforcing theright to be forgotten in complex data driven environments.

Link: [L1] https://arxiv.org/abs/1708.01104

References:

[1] A. Esteva, et al.: “Dermatologist-level classification ofskin cancer with deep neural networks”, Nature, 542,(7639), 115-118, 2017, doi:10.1038/nature21056.

[2] A. Holzinger, et al.: “A glass-box interactive machinelearning approach for solving NP-hard problems withthe human-in-the-loop”, arXiv:1708.01104.

[3] B. Malle, P. Kieseberg, S. Schrittwieser, A. Holzinger:“Privacy Aware Machine Learning and the ‘Right to beForgotten’”, ERCIM News 107, (3), 22-23, 2016.

Please contact:

Katharina Holzinger, SBA Research, Vienna, [email protected]

43

Figure�1:�AI�shows�impressive�success,�still�having�difficulty�in�generating�underlying

explanatory�structures.


formal Methods

for the Railway Sector

by Maurice ter Beek, Alessandro Fantechi, AlessioFerrari, Stefania Gnesi (ISTI-CNR, Italy), and RiccardoScopigno (ISMB, Italy)

Researchers from the Formal Methods and Tools group

of ISTI-CNR are working on a review and assessment of

the main formal modelling and verification languages

and tools used in the railway domain, with the aim of

evaluating the actual applicability of the most promising

ones to a moving block signalling system model provided

by an industrial partner. The research is being conducted

in the context of the H2020 Shift2Rail project ASTRail.

Compared with other transport sectors, the railway sector isnotoriously cautious about adopting technological innova-tions. This is commonly attributed to the sector’s robustsafety requirements. An example is smart route planning:while GNSS-based positioning systems have been in use forquite some time now in the avionics and automotive sectorsto provide accurate positioning and smart route planning, thecurrent train separation system is still based on fixed blocks –a block being the section of the track between two fixedpoints. In signalling systems, blocks start and end at signals,with their lengths designed to allow trains to operate as fre-quently as necessary (i.e., ranging from many kilometres forsecondary tracks to a few hundred metres for a busy com-muter line). The block sizes are determined based on param-eters such as the line’s speed limit, the train’s speed, thebraking characteristics of trains, sighting and reaction timeof drivers, etc. However, the faster trains are allowed to run,

the longer the braking distance and the longer the blocksneed to be, thus decreasing the line’s capacity. This isbecause stringent safety requirements impose the length offixed blocks to be based on the worst-case braking distance,regardless of the actual speed of the train.

With a moving block signalling system, in contrast, a safezone around the moving train can be computed, thus opti-mising the line’s exploitation (Figure 1). For this solution towork, it requires the precise absolute location, speed anddirection of each train, to be determined by a combination ofsensors: active and passive markers along the track, as wellas trainborne speedometers. One of the current challenges inthe railway sector is to make moving block signalling sys-tems as effective and precise as possible, including GNSSand leveraging on an integrated solution for signal outages(think, e.g., of tunnels) and the problem of multipaths [1].This is one of the main topics addressed by the projectASTRail: SAtellite-based Signalling and AutomationSysTems on Railways along with Formal Method andMoving Block Validation (Figure 2). A subsequent aim of theproject is to study the possibility of deploying the resultingprecise and reliable train localisation to improve automaticdriving technologies in the railway sector.

We are currently reviewing and assessing the main formalmodelling and verification languages and tools used in therailway domain in order to identify the ones that are mostmature for application in the railway industry [2] [3]. This isdone by combining a scientific literature review with inter-views with stakeholders of the railway sector. This willshortly result in a classification and ranking, from which wewill select a set of languages and tools to be adopted in all therelevant development stages of the systems addressed byASTRail. In particular, we will formalise and validate a


Figure�1:�Safe�braking�distance�between

trains�in�fixed�block�and�moving�block

signalling�systems�(Image�courtesy�of

Israel.abad/Wikimedia�Commons

distributed�under�the�CC�BY-SA�3.0

license).

Figure�2:�Illustrative�summary

of�ASTRail’s�objectives.


moving block model developed by SIRTI, which is anindustry leader in the design, production, installation andmaintenance of railway signalling systems.

ASTRail will run until August 2019 and is coordinated byRiccardo Scopigno from the Istituto Superiore Mario Boellasulle Tecnologie dell’Informazione e delle Telecomuni-cazioni (ISMB, Italy). Other partners are SIRTI S.p.A.(Italy), Ardanuy Ingeniería S.A. (Spain), École Nationale del’Aviation Civile (ENAC, France), Union des IndustriesFerroviaires Européennes (UNIFE, Belgium) and ISTI-CNR(Italy).

Link:

ASTRail: http://www.astrail.eu/

References:

[1] F. Rispoli, et al.: “Recent Progress in Application ofGNSS and Advanced Communications for Railway Sig-naling”, in 23rd Intl. Conf. Radioelektronika, IEEE,2013, 13-22. DOI: 10.1109/RadioElek.2013.6530882

[2] A. Fantechi, W. Fokkink, and A. Morzenti: “SomeTrends in Formal Methods Applications to Railway Sig-naling”, Chapter 4 in Formal Methods for IndustrialCritical Systems: A Survey of Applications (S. Gnesiand T. Margaria, eds.). John Wiley & Sons, 2013, 61-84. DOI: 10.1002/9781118459898.ch4

[3] A. Fantechi: “Twenty-Five Years of Formal Methodsand Railways: What Next?”, in Software Engineeringand Formal Methods, LNCS, vol. 8368. Springer, 2013,167-183. DOI: 10.1007/978-3-319-05032-4_13

Please contact:

Stefania Gnesi, ISTI-CNR, [email protected]

45

the Impacts of Low-

Quality training data

on Information Extraction

from Clinical Reports

by Diego Marcheggiani (University of Amsterdam) andFabrizio Sebastiani (CNR)

In a joint effort between the University of Amsterdam

and ISTI-CNR, researchers have studied the negative

impact that low-quality training data (i.e., training data

annotated by non-authoritative assessors) has on

information extraction (IE) accuracy.

Information Extraction (IE) is the task of designing softwareartifacts capable of extracting, from informal and unstructuredtexts, mentions of particular concepts, such as the names ofpeople, organisations and locations – where the task usuallygoes by the name of “named entity extraction”. Domain-spe-cific concepts, such as drug names, or drug dosages, or com-plex descriptions of prognoses, are also examples.

Many IE systems are based on supervised learning, i.e., relyon training an information extractor with texts where men-tions of the concepts of interest have been manually“labelled” (i.e., annotated via a markup language); in otherwords, the IE system learns to identify mentions of conceptsby analysing what manually identified mentions of the sameconcepts look like.

It seems intuitive that the quality of the manually assignedlabels (i.e., whether the manually identified portions of textare indeed mentions of the concept of interest, and whethertheir starting points and ending points have been identifiedprecisely) has a direct impact on the accuracy of the extractorthat results from the training. In other words, one wouldexpect that learning from inaccurate manual labels will lead

Figure�1:

Screenshot

displaying�a

radiological

report�annotated

according�to�the

concepts�of

interest.

to inaccurate automatic extraction, according to the familiar“garbage in, garbage out” principle.

However, the real extent to which inaccurately labelledtraining data impact on the accuracy of the resulting IEsystem, has seldom (if ever) been tested. Knowing how muchaccuracy we are going to lose by deploying low-quality labelsis important, because low-quality labels are a reality in manyreal-world situations. Labels may be low quality, for example,when the manual labelling has been performed by “turkers”(i.e., annotators recruited via Mechanical Turk or othercrowdsourcing platforms), or by junior staff or interns, orwhen it is old and outdated (so that the training data are nolonger representative of the data that the information extractorwill receive as input). What these situations share in commonis that the training data were manually labelled by one ormore “non-authoritative” annotators, i.e., by someone dif-ferent from the (“authoritative”) person who, in an ideal situa-tion (i.e., one where there are no time / cost / availability con-straints), would have annotated it.

In this work, the authors perform a systematic study of theextent to which low-quality training data negatively impactson the accuracy of the resulting information extractors. Thestudy is carried out by testing how accuracy deteriorateswhen training and test set have been annotated by two dif-ferent assessors. Naturally enough, the assessor who anno-tates the test data is taken to be the “authoritative” annotator(since accuracy is tested according to her/his judgment),while the one who annotates the training data is taken to bethe “non-authoritative” one. The study is carried out byapplying widely used “sequence learning” algorithms (eitherConditional Random Fields or Hidden Markov SupportVector Machines) on a doubly annotated dataset (i.e., adataset in which each document has independently beenannotated by the same two assessors) of radiological reports.Such reports, and clinical reports in general, are generated byclinicians during everyday practice, and are thus a chal-lenging type of text, since they tend to be formulated ininformal language and are usually fraught with typos, idio-syncratic abbreviations, and other types of deviations fromlinguistic orthodoxy. The fact that the dataset is doubly anno-tated allows the systematic comparison between high-qualitysettings (training set and test set annotated by annotator A)and low-quality settings (training set annotated by annotatorNA and test set annotated by annotator A), thereby allowingprecise quantification of the difference in accuracy betweenthe two settings.

Link:

http://nmis.isti.cnr.it/sebastiani/Publications/JDIQ2017.pdf

References:

[1] A. Esuli, D. Marcheggiani, F. Sebastiani: “AnEnhanced CRFs-based System for Information Extrac-tion from Radiology Reports”, Journal of BiomedicalInformatics 46, 3 (2013), 425–435.

[2] W. Webber, J. Pickens: “Assessor disagreement and textclassifier accuracy”, in Proc. of SIGIR 2013,929–932, 2013.

Please contact:

Fabrizio Sebastiani, ISTI-CNR, Italy+39 050 6212892 [email protected]

4th International Indoor

Positioning and Indoor

Navigation Competition

The fourth international Indoor Positioning and Indoor

Navigation (IPIN) competition, seventh in the EvAAL

series, was hosted by the international IPIN conference

in Sapporo, Japan, on 17 September 2017.

The aim of the competition is to measure the performance ofindoor localisation systems, that are usable in offices, hospi-tals or other big buildings like warehouses. The 2017 editionhas attracted 28 teams and allowed participants to test theirlocalisation solutions with rigorous procedures inside thetwo-floor structure of the Conference Hall of HokkaidoUniversity.

The competition ended with the awarding of four 150.000¥(1.100€) prizes:• smartphone-based (Chan Gook Park, Seoul National Uni-

versity, Corea)• dead reckoning (Chuanhua Lu, Kyushu University, Japan)• offline smartphone-based (Adriano Moreira, University of

Minho, Portugal)• offline PDR warehouse picking (Yoshihiro Ito, KDDI

R&D Laboratories Inc., Japan).

Prizes were awarded by the official sponsors of the competi-tion KICS , ETRI, TOPCON, and PDR Benchmark, respec-tively.

Francesco Potortì, Antonino Crivello and Filippo Palumbo,researchers at the WNLab group of CNR-ISTI at Pisa, wereamong the main organisers.

More information on the international competition, includ-ing complete results, photos and comments athttp://evaal.aaloa.org


46 ERCIM NEWS 112 January 2018

IPIN�participants.�


Events

Joint 22nd International

Workshop on formal Methods

for Industrial Critical Systems

and 17th International Workshop

on Automated verification of

Critical Systems

by Ana Cavalcanti (University of York), Laure Petrucci(LIPN, CNRS & Université Paris 13) and CristinaSeceleanu (Mälardalen University)

The yearly workshop of the ERCIM Working Group on

Formal Methods for Industrial Critical Systems (FMICS)

was organised as a joint event together with the

workshop on Automated Verification of Critical Systems

(AVoCS). The resulting FMICS-AVoCS 2017 workshop took

place on 18-20 September in Turin, hosted by the

University of Turin.

The aim of the FMICS workshop series is to provide a forumfor researchers interested in the development and applicationof formal methods in industry. It strives to promote researchand development for the improvement of formal methodsand tools for industrial applications. The aim of the AVoCSworkshop series is to contribute to the interaction andexchange of ideas among members of the international com-munity on tools and techniques for the verification of criticalsystems.

The workshop was chaired by Laure Petrucci (LIPN, CNRS& Université Paris 13, France) and Cristina Seceleanu(Mälardalen University, Sweden). A special track on “Formalmethods for mobile and autonomous robots” took placewithin the event, and was chaired by Ana Cavalcanti(University of York, UK). The workshop attracted 30 partici-pants from eleven countries.

A total of thirty papers were submitted, including eightspecifically for the special track. Fourteen of them wereaccepted (including four for the special track).

The programme also included two excellent invited keynotelectures: “Replacing store buffers by load buffers in totalstore ordering” by Parosh Abdulla (Uppsala University,Sweden) and “Towards formal apps: turning formal methodsinto verification techniques that make the difference in prac-tice” by Kerstin Eder (University of Bristol, UK). Moreover,a half-day tutorial took place: “DIME: Model-based genera-tion of running web applications” by Tiziana Margaria(University of Limerick & Lero, Ireland) and PhilipZweihoff (TU Dortmund, Germany).

The presentations were of extremely good quality. The pro-gramme committee awarded two best papers: “A unified for-malism for monoprocessor schedulability analysis underuncertainty” by Etienne André, and “Formalising the Dezynemodelling language in mCRL2” by Rutger van Beusekom,Jan Friso Groote, Paul Hoogendijk, Rob Howe, Wieger

Wesselink, Rob Wieringa and Tim Willemse. An excellentprogramme together with a nice weather contributed to thesuccess of the workshops.

We gratefully acknowledge the support of Springer for pub-lishing the workshop’s proceedings and sponsoring the bestpapers, EasyChair for assisting us in managing the completeprocess from submission to proceedings, as well as ERCIMand EASST.

The proceedings of FMICS-AVoCS 2017 have been pub-lished by Springer as volume 10471 of their LNCS series.

Selected papers are proposed for special issues of the inter-national journals Software Tools for Technology Transfer(STTT) and Science of Computer Programming (SCP).

Links:

FMICS-AVoCS 2017:http://www.es.mdh.se/conferences/fmics-avocs-2017/FMICS Working Group: http://fmics.inria.fr/

Reference:

Laure Petrucci, Cristina Seceleanu and Ana Cavalcanti(eds.): “Critical Systems: Formal Methods and AutomatedVerification -Joint 22nd International Workshop on Formal Methods forIndustrial Critical Systems and 17th International Work-shop on Automated Verification of Critical Systems(FMICS-AVoCS’17)”, Turin, Italy, 18–20 September 2017,Lecture notes in Computer Science 10471, Springer, 2017.https://doi.org/10.1007/978-3-319-67113-0

Please contact:

Ana Cavalcanti, University of [email protected]

Laure Petrucci, LIPN, CNRS & Université Paris [email protected]

Cristina Seceleanu, Mälardalen [email protected]

Best�paper�award�winners�Etienne�André�(left)�and�Tim�Willemse.�

visual Heritage 2018

Vienna, 12-15 November 2018

With the aim of continuing the suc-cessful experience of Digital Heritage2013 (Marseille, France) and DigitalHeritage 2015 (Granada, Spain), thenext edition of the EurographicsGraphics and Cultural Heritage (EGGCH) Symposium will be organized incooperation with CHNT (CulturalHeritage and New Technologies) inVienna. The aim of this federated eventis again to bring different communitiesin the same venue, to share experiencesand discuss methodologies concerningdigital visual media and their use in thecontext of cultural heritage applica-tions.

We are looking for a wide participationof our community, as well as the collab-oration and inclusion of other structuredcommunities and events, since VisualHeritage 2018 is aimed as an open circle(please contact the organizers at [email protected] if you are anorganization/community interested injoining us).

The 2018 edition will be a special event,since 2018 has been declared by theEuropean Commission the "EuropeanYear of Cultural Heritage". The event inVienna will also take place during theAustrian term as President of the EC.Therefore, VH2018 will be an idealcontext for discussing European poli-cies on digital heritage and digitalhumanities.

Visual Heritage 2018 will have inde-pendent call of papers for each eventcontributing to the program. The EGGCH 2018 program will be based on acall for papers (full and short papers)with deadline June 20th, 2018

Paper submissionMore details on the EG GCH scientifictopics, instructions for submitters, callfor paper dates, and the selection proce-dure will be published soon on EG GCHVH2018 web page http://2018.visual-heritage.org

More information

http://www.chnt.at/


Events

IfIP Networking 2018

The IFIP Networking 2018 Conference(NETWORKING 2018), to be held inZurich, Switzerland, from 14-16 May2018 is the 17th event of the series, spon-sored by the IFIP Technical Committeeon Communication Systems (TC6).Accepted papers will be published in theIFIP Digital Library and submitted to theIEEE Xplore Digital Library.

The main objective of Networking 2018is to bring together members of the net-working community, from both aca-demia and industry, to discuss recentadvances in the broad and quickly-evolving fields of computer and commu-nication networks, to highlight keyissues, identify trends, and develop avision for future Internet technology,operation, and use.

Important Dates: • Abstract registration: January 2, 2018

(everywhere on earth) • Full paper submission:

January 7, 2018 • Acceptance notification:March 5, 2018 • Camera-ready papers: March 23, 2018

Link:

http://networking.ifip.org/2018/

Web Science

Conference 2018

Amsterdam, 27-31 May 2018

The 10th International ACM Conferenceon Web Science in 2018 (WebSci’18) isa unique conference where a multitudeof disciplines converge in a creative andcritical dialogue with the aim of under-standing the Web and its impacts. Theconference brings together researchersfrom multiple disciplines, like computerscience, sociology, economics, informa-tion science, anthropology and psy-chology. Web Science is the emergentstudy of the people and technologies,applications, processes and practicesthat shape and are shaped by the WorldWide Web.

Keynote speakers include: Tim Berners-Lee, José van Dijck and John Domingue

More information:

https://websci18.webscience.org

10th International

Conference of the

ERCIM Working Group

on Computational

and Methodological

Statistics

The 10th International Conference of theERCIM WG on Computational andMethodological Statistics (CMStatistics2017) took place at the Senate Houseand Birkbeck, University of London,UK, 16-18 December 2017. Tutorialswere given on Friday 15th of December2017 and the COST IC1408 CRoNoSWinter Course on Copula-based model-ling with R took place the 13-14December 2017. The conference tookplace jointly with the 11th InternationalConference on Computational andFinancial Econometrics (CFE 2017).

This annual conference has become aleading joint international meeting at theinterface of statistics, econometrics,empirical finance and computing. Theconference aims at bringing togetherresearchers and practitioners to discussrecent developments in computationalmethods for economics, finance, andstatistics. The CFE-CMStatistics2017programme comprised of 375 sessions,5 plenary talks and over 1550 presenta-tions. There were about 1700 partici-pants. This was the biggest meeting ofthe conference series in terms of numberof participants and presentations. Thegrowth of the conference in terms of sizeand quality makes it undoubtedly one ofthe most important international scien-tific events in the field.

Link:

http://www.cmstatistics.org/

Please contact:

[email protected]

Postdoc Position in the Research Project

“Approximation Algorithms, Quantum Information

and Semidefinite optimization”

Centrum Wiskunde & Informatica (CWI) has a vacancy in the Network & Optimization researchgroup for a talented Postdoc in the research project "Approximation Algorithms, QuantumInformation and Semidefinite Optimization”.

Job descriptionThe research project “Approximation algorithms, quantum information and semidefinite opti-mization” aims to explore the limits of efficient computation within classical and quantum com-puting, using semidefinite optimization as a main unifying tool. The position involves researchinto the mathematical and computer science aspects of approximation algorithms for discreteoptimization, quantum entanglement in communication, and complexity of fundamental prob-lems in classical and quantum computing. The research will be supervised by Prof. MoniqueLaurent from the CWI Networks & Optimization research group, in collaboration with Prof.Ronald de Wolf from the CWI Algorithms & Complexity research group, and Prof. Nikhil Bansalfrom the department of mathematics and computer science of the Technical UniversityEindhoven. The position is funded through an NWO-TOP grant.

RequirementsCandidates are required to have a completed PhD in mathematics and/or computer science, and anexcellent research track-record. The ideal candidate will have a strong mathematical backgroundand knowledge in several of the following topics: algorithms, complexity, algebra, combinatorialoptimization, semidefinite optimization, quantum information, communication and informationtheory. The candidate should also have a taste in interdisciplinary research at the frontier betweenmathematics and computer science. Further needed qualifications for candidates include provenresearch talent and good academic writing and presentation skills. Candidates are expected tohave an excellent command of English.

Terms and conditionsThe terms of employment are in accordance with the Dutch Collective Labour Agreement forResearch Centres (“CAO-onderzoeksinstellingen”). The gross monthly salary for an employee ona full time basis, depending on relevant work experience, ranges from € 3,409 to € 4,154.Employees are also entitled to a holiday allowance of 8% of the gross annual salary and a year-end bonus of 8.33%. CWI offers attractive working conditions, including flexible scheduling. Theappointment will be for a period of one year, starting as soon as possible before September 2018.

Application Applications can be sent to [email protected]. We will take applications until getting our positionfilled. All applications should include a detailed CV, a motivation letter describing research inter-ests and reasons for applying to this project, a research statement, and a short description of thePhD dissertation or of the most relevant publications. The applicant should provide names of upto three scientists who are acquainted with their previous academic performance and can writereference letters.

More information• Position announcement: https://kwz.me/hBF• About the Networks and Optimization at CWI:

https://www.cwi.nl/research/groups/networks-and-optimization • For CWI, please visit https://www.cwi.nl or watch our video “A Fundamental Difference”

about working at CWI (https://www.cwi.nl/general/a-fundamental-difference).• About the vacancy, please contact Prof. Monique Laurent, [email protected].


CIdoC Conference

2018

Heraklion, Crete, Greece, 29September – 4 October 2018

Conference theme: Provenance of

Knowledge

Participants in the event are invited toconsider the theme of this conference,‘Provenance of Knowledge’, both in itsbroad sense and in its technical detail.Today, museum professionals have totake into account as much the traditionalresearch and documentation of the his-tory of an object as the contemporarydigital interpretation in the sense of thehistory of the transformations of a dig-ital object. The conference invites par-ticipants to share their understanding ofthe shifting interpretations and uses ofthe notion of ‘provenance’:• In what ways can it, or can it not,

facilitate grounding in ‘knowledge’? • What is the role of the museum as

mediator of cultural heritage in facili-tating and establishing such prove-nance and knowledge?

Conference topics:• Documentation & interdisciplinarity• Object documentation and analytical

resources • Provenance of materials and tech-

niques• Field research and object documenta-

tion• Object documentation and archival

resources• Oral tradition and witnessing infor-

mation & connection with objects• Methods of knowledge verification

and documentation of knowledgerevision (“subjective” , “objective”and other forms of evidence)

• Documentation for target groups (e.g.special needs, etc)

• Object information as historicalsource / evidence (ethics of prove-nance of information)

More information:www.cidoc2018.com

ICOMinternational committee for documentation

2018ConferenceCIDOC20182018201820182018

CIDOC2018

CIDOC2018

international ICOM

2018ConferenceConferenceConference documentation

committee for documentationcommittee for

ERCIM

Membership

After having successfully grown tobecome one of the most recognized ICTSocieties in Europe, ERCIM has openedmembership to multiple member institutesper country. By joining ERCIM, yourresearch institution or university candirectly participate in ERCIM’s activitiesand contribute to the ERCIM members’common objectives playing a leading rolein Information and CommunicationTechnology in Europe:• Building a Europe-wide, open network

of centres of excellence in ICT andApplied Mathematics;

• Excelling in research and acting as abridge for ICT applications;

• Being internationally recognised both asa major representative organisation in itsfield and as a portal giving access to allrelevant ICT research groups in Europe;

• Liaising with other international organi-sations in its field;

• Promoting cooperation in research,technology transfer, innovation andtraining.

About ERCIM

ERCIM – the European ResearchConsortium for Informatics andMathematics – aims to foster collaborativework within the European research com-munity and to increase cooperation withEuropean industry. Founded in 1989,ERCIM currently includes 15 leadingresearch establishments from 14 Europeancountries. ERCIM is able to undertake con-sultancy, development and educationalprojects on any subject related to its field ofactivity.

ERCIM members are centres of excellenceacross Europe. ERCIM is internationallyrecognized as a major representativeorganization in its field. ERCIM providesaccess to all major InformationCommunication Technology researchgroups in Europe and has established anextensive program in the fields of science,strategy, human capital and outreach.ERCIM publishes ERCIM News, a quar-terly high quality magazine and deliversannually the Cor Baayen Award to out-standing young researchers in computerscience or applied mathematics. ERCIMalso hosts the European branch of theWorld Wide Web Consortium (W3C).

benefits of Membership

As members of ERCIM AISBL, institutions benefit from: • International recognition as a leading centre for ICT R&D, as member of the

ERCIM European-wide network of centres of excellence;• More influence on European and national government R&D strategy in ICT.

ERCIM members team up to speak with a common voice and produce strategicreports to shape the European research agenda;

• Privileged access to standardisation bodies, such as the W3C which is hosted byERCIM, and to other bodies with which ERCIM has also established strategiccooperation. These include ETSI, the European Mathematical Society and Infor-matics Europe;

• Invitations to join projects of strategic importance;• Establishing personal contacts with executives of leading European research insti-

tutes during the bi-annual ERCIM meetings; • Invitations to join committees and boards developing ICT strategy nationally and

internationally;• Excellent networking possibilities with more than 10,000 research colleagues

across Europe. ERCIM’s mobility activities, such as the fellowship programme,leverage scientific cooperation and excellence;

• Professional development of staff including international recognition;• Publicity through the ERCIM website and ERCIM News, the widely read quarter-

ly magazine.

How to become a Member

• Prospective members must be outstanding research institutions (including univer-sities) within their country;

• Applicants should address a request to the ERCIM Office. The application shouldinlcude:

• Name and address of the institution;• Short description of the institution’s activities;• Staff (full time equivalent) relevant to ERCIM’s fields of activity;• Number of European projects in which the institution is currently involved;• Name of the representative and a deputy.

• Membership applications will be reviewed by an internal board and may includean on-site visit;

• The decision on admission of new members is made by the General Assembly ofthe Association, in accordance with the procedure defined in the Bylaws(http://kwz.me/U7), and notified in writing by the Secretary to the applicant;

• Admission becomes effective upon payment of the appropriate membership fee ineach year of membership;

• Membership is renewable as long as the criteria for excellence in research and anactive participation in the ERCIM community, cooperating for excellence, are met.

Please contact the ERCIM Office: [email protected]

“Through a long history of successful research collaborations

in projects and working groups and a highly-selective mobility

programme, ERCIM has managed to become the premier net-

work of ICT research institutions in Europe. ERCIM has a consis-

tent presence in EU funded research programmes conducting

and promoting high-end research with European and global

impact. It has a strong position in advising at the research pol-

icy level and contributes significantly to the shaping of EC

framework programmes. ERCIM provides a unique pool of

research resources within Europe fostering both the career

development of young researchers and the synergies among

established groups. Membership is a privilege.”Dimitris Plexousakis, ICS-FORTH, ERCIM AISBL Board


CWI hosts EIt digital’s New Innovation Space

CWI houses a new innovation space of partner EIT Digital, which was opened on 2November. With this new location in the financial heart of the Netherlands, EITDigital wants to boost the development of FinTech, together with its partners in theNetherlands and Europe.

Jos Baeten (CWI

Director) and

Patrick Essers

(Node Director

EIT Digital in the

Netherlands)

open EIT

Digital’s new

Innovation Space

at CWI.

The new satellite location offers EIT Digital a strong base to help FinTech companiesand to develop new digital FinTech initiatives. The reason for opening a second EITDigital location in the Netherlands is a strongly felt wish of the EIT Digital’s partnersBright Cape, CWI, ING Bank, and TNO who cooperate in various FinTech innova-tion activities within the pan-European ecosystem of EIT Digital. The Municipalityof Amsterdam supports this and CWI is offering the new location.

EIT Digital is a leading European innovation and education organization. Its missionis to foster digital technology innovation and entrepreneurial talent for economicgrowth and quality of life in Europe. It brings together entrepreneurs from a partner-ship of over 130 European corporations, SMEs, start-ups, universities and researchinstitutes.

In brief

Google Awards Grant

for fake News

detection to foRtH

and University of

Cyprus

As part of its Digital News Initiative(DNI), Google announced a €150 millioninnovation fund that supports innovationin Digital News Journalism. In its mostrecent round of funding, Google sup-ported “Check-it: Visualizing fake newson social media”, a collaborative projectbetween FORTH and University ofCyprus.

Check-it empowers users with the toolsthey need in order to check whether thestories they read on-line are fake or not.In this way (i) users will be able to see ifwhat they read is fake, and (ii) they maybe reluctant to forward news which areknown to be fake.

The main problem that this project triesto address is that when users view newson social media they usually have little,if any, means to decide whether theinformation they see is real or fake.Although it is true that some socialmedia allow users to report any fakenews they see, most of the social mediaout there do not provide such function-ality. Therefore, they place all theburden of deciding whether a story isfake or real on the end user. In thisproject we propose to change the waypeople consume digital news byempowering users with an automatedability to “check” whether a story isfake or not. Towards this end, theproject extends the web browser capa-bilities with a “check it” button (abrowser plug-in) that will enable usersto check whether a story is true or fake.

For more information, please contact Evangelos Markatos([email protected]), Marios Dikaiakos([email protected]) and G Pallis ([email protected])

CWI merges with NWo Institutes organisation

From 1 January 2018, the ERCIM member CWI has been merged with the NWOInstitutes Organisation, NWO-I. The other research institutes in the Netherlands thatjoined the recent merger are ASTRON, NIOZ, NSCR and SRON. The Dutch insti-tutes AMOLF, ARCNL, DIFFER and Nikhef were already part of NWO-I.

In 2015, the Netherlands Organisation for Scientific Research (NWO) started a tran-sition to a new organizational structure. The new NWO is intended to be more flex-ible, more effective, and more focused on collaboration, which means it will be in abetter position to respond to developments in science and society. In the new NWOstructure there is a clear distinction between a granting organization, NWO, and aseparate institutes organization: the Netherlands Foundation of Scientific ResearchInstitutes, or NWO-I for short.

After the merger, CWI will continue to fulfil its mission to conduct pioneeringresearch in mathematics and computer science, generating new knowledge in thesefields and conveying it to industry and to society at large. .The general director ofCWI, Jos Baeten, continues to be responsible for the CWI mission, for the day to daymanagement of the institute, for the appointment of all CWI personnel and for theimplementation of (research) policies. The foundation board of NWO-I will adviseand monitor Baeten.

For more information about the merger, visit the NWO-I website: https://www.nwo-i.nl/en/

ERCIM is the European Host of the World Wide Web Consortium.

Institut National de Recherche en Informatique et en AutomatiqueB.P. 105, F-78153 Le Chesnay, Francehttp://www.inria.fr/

VTT Technical Research Centre of Finland LtdPO Box 1000FIN-02044 VTT, Finlandhttp://www.vttresearch.com

SBA Research gGmbHFavoritenstraße 16, 1040 Wienhttp://www.sba-research.org/

Norwegian University of Science and Technology Faculty of Information Technology, Mathematics and Electri-cal Engineering, N 7491 Trondheim, Norwayhttp://www.ntnu.no/

Universty of WarsawFaculty of Mathematics, Informatics and MechanicsBanacha 2, 02-097 Warsaw, Polandhttp://www.mimuw.edu.pl/

Consiglio Nazionale delle RicercheArea della Ricerca CNR di PisaVia G. Moruzzi 1, 56124 Pisa, Italyhttp://www.iit.cnr.it/

Centrum Wiskunde & InformaticaScience Park 123, NL-1098 XG Amsterdam, The Netherlandshttp://www.cwi.nl/

Foundation for Research and Technology – HellasInstitute of Computer ScienceP.O. Box 1385, GR-71110 Heraklion, Crete, Greecehttp://www.ics.forth.gr/FORTH

Fonds National de la Recherche6, rue Antoine de Saint-Exupéry, B.P. 1777L-1017 Luxembourg-Kirchberghttp://www.fnr.lu/

Fraunhofer ICT GroupAnna-Louisa-Karsch-Str. 210178 Berlin, Germanyhttp://www.iuk.fraunhofer.de/

RISE SICSBox 1263, SE-164 29 Kista, Swedenhttp://www.sics.se/

Magyar Tudományos AkadémiaSzámítástechnikai és Automatizálási Kutató IntézetP.O. Box 63, H-1518 Budapest, Hungaryhttp://www.sztaki.hu/

University of CyprusP.O. Box 205371678 Nicosia, Cyprushttp://www.cs.ucy.ac.cy/

Subscribe to ERCIM News and order back copies at http://ercim-news.ercim.eu/

ERCIM – the European Research Consortium for Informatics and Mathematics is an organisa-

tion dedicated to the advancement of European research and development in information tech-

nology and applied mathematics. Its member institutions aim to foster collaborative work with-

in the European research community and to increase co-operation with European industry.

INESCc/o INESC Porto, Campus da FEUP, Rua Dr. Roberto Frias, nº 378,4200-465 Porto, Portugal

I.S.I. – Industrial Systems InstitutePatras Science Park buildingPlatani, Patras, Greece, GR-26504 http://www.isi.gr/


Documents

SQpeciual thaeme ntum - ERCIM News 112 · SQpeciual thaeme ntum Computing Also in this issue: Research and Innovation: Computers that negotiate on our behalf. ERCIM News is the magazine