Bill Martin Mathematical Sciences and Computer Science Worcester Polytechnic Institute

  • View
    216

  • Download
    2

Embed Size (px)

Citation preview

  • Slide 1
  • Bill Martin Mathematical Sciences and Computer Science Worcester Polytechnic Institute
  • Slide 2
  • Many photos borrowed from the web (sources available on request) This talk focuses only on the combinatorics; there is a lot more activity that I wont talk about WPI is looking for graduate students and visiting faculty
  • Slide 3
  • Slide 4
  • Slide 5
  • Slide 6
  • Slide 7
  • Slide 8
  • Slide 9
  • Slide 10
  • Slide 11
  • Slide 12
  • Slide 13
  • Slide 14
  • Slide 15
  • Slide 16
  • Slide 17
  • Slide 18
  • Quadrature rules Numerical simulation Global optimization
  • Slide 19
  • Random Pseudo-random (should fool an observer) Quasi-Random: entirely deterministic, but has some statistical properties that a random set should have
  • Slide 20
  • Random (Monte Carlo) Lattice rules Latin hypercube sampling (T,M,S)-nets
  • Slide 21
  • A set N of N points inside [0,1) s An interval E = [0,a 1 )x[0,a 2 )x... x[0,a s ) should contain Vol(E) | N | of these points The star discrepancy of a set N of N points in [0,1) s is the supremum of | | N E| / N - Vol(E) | taken over all such intervals E. Call it D * ( N ) U
  • Slide 22
  • J. KoksmaE. Hlawka
  • Slide 23
  • For any given shape (d 1,d 2,...,d s ), the unit cube is partitioned into b m elementary intervals of this shape, each being a translate of every other.
  • Slide 24
  • Vienna, Austria 1980s
  • Slide 25
  • Harald Niederreiter Working on low discrepancy sequences, quasi-randomness, pseudo-random generators, applications to numerical analysis, coding theory, cryptography Expertise in finite fields and number theory
  • Slide 26
  • Niederreiter (1987), generalizing an idea of Sobol (1967)
  • Slide 27
  • Slide 28
  • Slide 29
  • Slide 30
  • Slide 31
  • Slide 32
  • Slide 33
  • Slide 34
  • Slide 35
  • Slide 36
  • Slide 37
  • Slide 38
  • Slide 39
  • Slide 40
  • Two MOLS(3) yield an orthogonal array of strength two
  • Slide 41
  • Replace alphabet by {0,1,,b-1} (here, base b=3)
  • Slide 42
  • Insert decimal points to obtain a (0,2,2)-net in base 3
  • Slide 43
  • (0,2,2)-net in base 3
  • Slide 44
  • Now fill in with cosets of the linear code
  • Slide 45
  • Vienna, Austria 1980s Madison, Wisconsin 1995
  • Slide 46
  • Mark Lawrence, Chief Risk Officer, Australia and New Zealand Banking Group
  • Slide 47
  • In an orthogonal array of strength t, all entries are chosen from some fixed alphabet {0,1,...,b-1} In any t columns, every possible t-tuple over the alphabet (there are q t of these) appears equally often So the total number of rows is.b t where is the replication number If this hold for a set of columns, then it also holds for all subsets of that set Now specify a partial order on the columns and require this only for lower ideals in this poset of size t or less
  • Slide 48
  • Vienna, Austria 1980s Salzburg, Austria 1995
  • Slide 49
  • Wolfgang Ch. Schmid and Gary Mullen Introduced OOA concept Proved equivalence to (T,M,S)-nets constructions bounds
  • Slide 50
  • Slide 51
  • 0 0 1 01 1 1 00 1 1 1 0 1 01 1 1 00 10 1 0 1 1 01 1 00 0 1 1 0
  • Slide 52
  • 0 0 0 0 0 11 0 11 1 1 1 0 10 1 1 1 1 10 0 11 0 1 1 1 01 0 00 1 0 0 1 11 1 10 0 1 1 0 01 1 01 0 0 0 1 0 1 1 0
  • Slide 53
  • There exists a (T,M,S)-net in base b If and only if there exists an OOA t, s, l, v) where s=S t=l=M-T v=b = b T
  • Slide 54
  • Slide 55
  • Vienna, Austria 1980s Singapore 1995
  • Slide 56
  • Harald Niederreiter and Chaoping Xing ( here pictured with Sang Lin) Global function fields with many rational places
  • Slide 57
  • For simplicity, assume q is a prime Let S = { p 1, p 2,..., p s } be a subset of F q (or PG(1,q) ) Fix k >= 0 and create one point for each polynomial f(x) in F q [x] of degree k or less In the i th coordinate position, take f(p i )/q + f (1) (p i )/q 2 +... + f (k) (p i )/q k+1 where f (j) denotes the j th derivative of f
  • Slide 58
  • To illustrate, lets take q = 5 k = 2 S = { 1, 2, 3} inside F 5 For example, the polynomial f(x) = 3 x 2 + 4 x has f (1) (x) = x + 4 and f (2) (x) = 1 This contributes the point in [0,1) 3 (.208,.048,.888 )
  • Slide 59
  • First 5 points (constant polys)
  • Slide 60
  • First 10 pts (constant &linear)
  • Slide 61
  • First 15 points (constant & linear)
  • Slide 62
  • First 20 points (constant & linear)
  • Slide 63
  • First 25 points (all const & lin)
  • Slide 64
  • First 50 points
  • Slide 65
  • First 75 points
  • Slide 66
  • First 100 points
  • Slide 67
  • All 125 points
  • Slide 68
  • All 125 points another viewpoint
  • Slide 69
  • Vienna, Austria 1980s Heidelberg, Germany 1995
  • Slide 70
  • Vienna, Austria 1980s Houghton, Michigan 1995
  • Slide 71
  • Yves Edel and Juergen Bierbrauer Digital nets from BCH codes ... and twisted BCH codes
  • Slide 72
  • Vienna, Austria 1980s Moscow, Russia 1995
  • Slide 73
  • M. Yu. Rosenbloom and Michael Tsfasman Codewords are matrices Errors affect entire tail of a row algebraic geometry codes Gilbert-Varshamov bound ... and more
  • Slide 74
  • Vienna, Austria 1980s
  • Slide 75
  • Auburn workshop in 1995 Reception at Pebble Hill Juergen Bierbrauer teaches me about (t,m,s)-nets over snacks Questions: Is there a linear programming bound for these things? Is there a MacWilliams-type theorem for duality?
  • Slide 76
  • Vienna, Austria 1980s
  • Slide 77
  • Slide 78
  • Michael Adams Completed dissertation at U. Wyoming under Bryan Shader Poset metrics for codes New constructions of nets Convincing argument that MacWilliams identities DONT exist
  • Slide 79
  • Vienna, Austria 1980s Winnipeg, Manitoba 1997
  • Slide 80
  • Vienna, Austria 1980s Winnipeg, March 1997
  • Slide 81
  • Vienna, Austria 1980s University of Manitoba
  • Slide 82
  • Vienna, Austria 1980s University of Nebraska
  • Slide 83
  • Slide 84
  • Slide 85
  • Doug Stinson and WJM Self-dual association scheme generalising the Hamming schemes Duality between codes and OOAs MacWilliams identities, LP bound
  • Slide 86
  • Slide 87
  • Slide 88
  • Vladimir Levenshtein BCC at Queen Mary & Westfield College (qmul) Look at this paper by Rosenbloom and Tsfasman
  • Slide 89
  • Slide 90
  • Slide 91
  • Slide 92
  • Slide 93
  • Slide 94
  • Slide 95
  • St. Petersburg, Russia 1999
  • Slide 96
  • Steven Dougherty and Maxim Skriganov
  • Slide 97
  • Skriganov and then Dougherty/Skriganov: independently re-discovered a lot of the above MDS codes for the m-metric MacWilliams identities bounds and constructions
  • Slide 98
  • Houghton, Michigan
  • Slide 99
  • Vienna, Austria 1980s Winnipeg, Manitoba 1997
  • Slide 100
  • Terry Visentin and WJM
  • Slide 101
  • Vienna, Austria 1980s Salzburg, Austria 1995
  • Slide 102
  • Wolfgang Ch. Schmid and Rudi Schurer Many contributions But also a comprehensive on-line table of parameters with links to literature
  • Slide 103
  • Slide 104
  • Thank You
  • Slide 105
  • Slide 106
  • Slide 107
  • Sga12345677890- qwery xcbaABKFQWFIOQWUFO: EIVNS U N n
  • Slide 108
  • TTTThis is the text I want BBBBUT THIS IS BETTER NNNNOW WE HAVE ANOTHER OPTION VVVVIENNA, AUSTRIA: MAY 1986
  • Slide 109