51
Hyperdimensional Cognitive Computing: A New Approach to Some Very Old Problems Simon D. Levy Washington & Lee University Lexington,Virginia, USA Tuesday, April 14, 2009

Hyperdimensional Cognitive Computing: A New Approach to ......Hyperdimensional Cognitive Computing: A New Approach to Some Very Old Problems Simon D. Levy Washington & Lee University

  • Upload
    others

  • View
    7

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Hyperdimensional Cognitive Computing: A New Approach to ......Hyperdimensional Cognitive Computing: A New Approach to Some Very Old Problems Simon D. Levy Washington & Lee University

Hyperdimensional Cognitive Computing: A New Approach to

Some Very Old Problems

Simon D. LevyWashington & Lee University

Lexington, Virginia, USA

Tuesday, April 14, 2009

Page 2: Hyperdimensional Cognitive Computing: A New Approach to ......Hyperdimensional Cognitive Computing: A New Approach to Some Very Old Problems Simon D. Levy Washington & Lee University

Inspiration(s)

[I]t turns out that we don’t think the way we think we think! ... The scientific evidence coming in all around us is clear: Symbolic conscious reasoning, which is extracted through protocol analysis from serial verbal introspection, is a myth.

− J. Pollack (2005)

[W]hat kinds of things suggested by the architecture of the brain, if we modeled them mathematically, could give some properties that we associate with mind?

− P. Kanerva (to appear)

Tuesday, April 14, 2009

Page 3: Hyperdimensional Cognitive Computing: A New Approach to ......Hyperdimensional Cognitive Computing: A New Approach to Some Very Old Problems Simon D. Levy Washington & Lee University

What is Mind?

Tuesday, April 14, 2009

Page 4: Hyperdimensional Cognitive Computing: A New Approach to ......Hyperdimensional Cognitive Computing: A New Approach to Some Very Old Problems Simon D. Levy Washington & Lee University

What is Mind?

Tuesday, April 14, 2009

Page 5: Hyperdimensional Cognitive Computing: A New Approach to ......Hyperdimensional Cognitive Computing: A New Approach to Some Very Old Problems Simon D. Levy Washington & Lee University

Tuesday, April 14, 2009

Page 6: Hyperdimensional Cognitive Computing: A New Approach to ......Hyperdimensional Cognitive Computing: A New Approach to Some Very Old Problems Simon D. Levy Washington & Lee University

Tuesday, April 14, 2009

Page 7: Hyperdimensional Cognitive Computing: A New Approach to ......Hyperdimensional Cognitive Computing: A New Approach to Some Very Old Problems Simon D. Levy Washington & Lee University

The Need for New Representational Principles• Ecological affordances (Gibson

1979); exploiting the environment (Clark 1998)

• Distributed/Connectionist Representations (PDP 1986)

• Holographic Representations (Gabor 1971; Plate 2003)

• Fractals / Attractors / Dynamical Systems (Tabor 2000; Levy & Pollack 2001)

Tuesday, April 14, 2009

Page 8: Hyperdimensional Cognitive Computing: A New Approach to ......Hyperdimensional Cognitive Computing: A New Approach to Some Very Old Problems Simon D. Levy Washington & Lee University

The Need for New Representational Principles• Ecological affordances (Gibson

1979); exploiting the environment (Clark 1998)

• Distributed/Connectionist Representations (PDP 1986)

• Holographic Representations (Gabor 1971; Plate 2003)

• Fractals / Attractors / Dynamical Systems (Tabor 2000; Levy & Pollack 2001)

Tuesday, April 14, 2009

Page 9: Hyperdimensional Cognitive Computing: A New Approach to ......Hyperdimensional Cognitive Computing: A New Approach to Some Very Old Problems Simon D. Levy Washington & Lee University

Pitfalls to Avoid

1. The “Short Circuit” (Localist Connectionist) Approachi) Traditional models of phenomenon X (language) use entities A,

B, C, ... (Noun Phrase, Phoneme, ...)

ii) We wish to model X in a more biologically realistic way.

iii) Therefore our model of X will have a neuron (pool) for A, one for B, one for C, etc.

Tuesday, April 14, 2009

Page 10: Hyperdimensional Cognitive Computing: A New Approach to ......Hyperdimensional Cognitive Computing: A New Approach to Some Very Old Problems Simon D. Levy Washington & Lee University

a.k.a. The Reese’s Peanut Butter Cup Model

Tuesday, April 14, 2009

Page 11: Hyperdimensional Cognitive Computing: A New Approach to ......Hyperdimensional Cognitive Computing: A New Approach to Some Very Old Problems Simon D. Levy Washington & Lee University

E.g. Neural Blackboard Model (van der Velde & de Kamps 2006)

Tuesday, April 14, 2009

Page 12: Hyperdimensional Cognitive Computing: A New Approach to ......Hyperdimensional Cognitive Computing: A New Approach to Some Very Old Problems Simon D. Levy Washington & Lee University

Benefits of Localism (Page 2000)

• Transparent (one node, one concept)

• Supports lateral inhibition / winner-takes all

Tuesday, April 14, 2009

Page 13: Hyperdimensional Cognitive Computing: A New Approach to ......Hyperdimensional Cognitive Computing: A New Approach to Some Very Old Problems Simon D. Levy Washington & Lee University

Lateral Inhibition (WTA)

A B C

L1

L2

Tuesday, April 14, 2009

Page 14: Hyperdimensional Cognitive Computing: A New Approach to ......Hyperdimensional Cognitive Computing: A New Approach to Some Very Old Problems Simon D. Levy Washington & Lee University

Problems with Localism

• Philosophical problem: “fresh coat of paint on old rotting theories” (MacLennan 1991): what new insights does “neuro-X” provide?

• Engineering problem: need to recruit new hardware for each new concept/combination leads to combinatorial explosion (Stewart & Eliasmith 2008)

Tuesday, April 14, 2009

Page 15: Hyperdimensional Cognitive Computing: A New Approach to ......Hyperdimensional Cognitive Computing: A New Approach to Some Very Old Problems Simon D. Levy Washington & Lee University

The Appeal of Distributed

Representations(Rumelhart &

McClelland 1986)

Tuesday, April 14, 2009

Page 16: Hyperdimensional Cognitive Computing: A New Approach to ......Hyperdimensional Cognitive Computing: A New Approach to Some Very Old Problems Simon D. Levy Washington & Lee University

WALK

WALKED

Tuesday, April 14, 2009

Page 17: Hyperdimensional Cognitive Computing: A New Approach to ......Hyperdimensional Cognitive Computing: A New Approach to Some Very Old Problems Simon D. Levy Washington & Lee University

ROAR

ROARED

Tuesday, April 14, 2009

Page 18: Hyperdimensional Cognitive Computing: A New Approach to ......Hyperdimensional Cognitive Computing: A New Approach to Some Very Old Problems Simon D. Levy Washington & Lee University

SPEAK

SPOKE

Tuesday, April 14, 2009

Page 19: Hyperdimensional Cognitive Computing: A New Approach to ......Hyperdimensional Cognitive Computing: A New Approach to Some Very Old Problems Simon D. Levy Washington & Lee University

GO

WENT

Tuesday, April 14, 2009

Page 20: Hyperdimensional Cognitive Computing: A New Approach to ......Hyperdimensional Cognitive Computing: A New Approach to Some Very Old Problems Simon D. Levy Washington & Lee University

ignores(mary, john)

Mary won’t give John the time of day.

Tuesday, April 14, 2009

Page 21: Hyperdimensional Cognitive Computing: A New Approach to ......Hyperdimensional Cognitive Computing: A New Approach to Some Very Old Problems Simon D. Levy Washington & Lee University

Challenges (Jackendoff 2002)

Tuesday, April 14, 2009

Page 22: Hyperdimensional Cognitive Computing: A New Approach to ......Hyperdimensional Cognitive Computing: A New Approach to Some Very Old Problems Simon D. Levy Washington & Lee University

I. The Binding Problem

+

? ? ? ?Tuesday, April 14, 2009

Page 23: Hyperdimensional Cognitive Computing: A New Approach to ......Hyperdimensional Cognitive Computing: A New Approach to Some Very Old Problems Simon D. Levy Washington & Lee University

II. The Problem of Two

+

? ? ?Tuesday, April 14, 2009

Page 24: Hyperdimensional Cognitive Computing: A New Approach to ......Hyperdimensional Cognitive Computing: A New Approach to Some Very Old Problems Simon D. Levy Washington & Lee University

III. The Problem of Variables

ignores(X, Y)

X won’t give Y the time of day.

Tuesday, April 14, 2009

Page 25: Hyperdimensional Cognitive Computing: A New Approach to ......Hyperdimensional Cognitive Computing: A New Approach to Some Very Old Problems Simon D. Levy Washington & Lee University

IV. Binding in Working Memory vs. Long-Term Memory

ignores(X, Y) X won’t give Y the time of day.

Long-term memory (idiosyncratic):

transitive-action(X, Y, Z) X Verb Y Z

Working memory (“rule-based”):

Mary gave John an apple.

Tuesday, April 14, 2009

Page 26: Hyperdimensional Cognitive Computing: A New Approach to ......Hyperdimensional Cognitive Computing: A New Approach to Some Very Old Problems Simon D. Levy Washington & Lee University

Vector Symbolic Architectures

(Plate 1991; Kanerva 1994; Gayler 1998)

Tuesday, April 14, 2009

Page 27: Hyperdimensional Cognitive Computing: A New Approach to ......Hyperdimensional Cognitive Computing: A New Approach to Some Very Old Problems Simon D. Levy Washington & Lee University

Tensor Product Binding(Smolensky 1990)

Tuesday, April 14, 2009

Page 28: Hyperdimensional Cognitive Computing: A New Approach to ......Hyperdimensional Cognitive Computing: A New Approach to Some Very Old Problems Simon D. Levy Washington & Lee University

Binding

Tuesday, April 14, 2009

Page 29: Hyperdimensional Cognitive Computing: A New Approach to ......Hyperdimensional Cognitive Computing: A New Approach to Some Very Old Problems Simon D. Levy Washington & Lee University

Bundling

+ =

Tuesday, April 14, 2009

Page 30: Hyperdimensional Cognitive Computing: A New Approach to ......Hyperdimensional Cognitive Computing: A New Approach to Some Very Old Problems Simon D. Levy Washington & Lee University

Unbinding (query)

Tuesday, April 14, 2009

Page 31: Hyperdimensional Cognitive Computing: A New Approach to ......Hyperdimensional Cognitive Computing: A New Approach to Some Very Old Problems Simon D. Levy Washington & Lee University

Lossy

Tuesday, April 14, 2009

Page 32: Hyperdimensional Cognitive Computing: A New Approach to ......Hyperdimensional Cognitive Computing: A New Approach to Some Very Old Problems Simon D. Levy Washington & Lee University

Lossy

Tuesday, April 14, 2009

Page 33: Hyperdimensional Cognitive Computing: A New Approach to ......Hyperdimensional Cognitive Computing: A New Approach to Some Very Old Problems Simon D. Levy Washington & Lee University

Cleanup

Hebbian / Hopfield /

Attractor Net

Tuesday, April 14, 2009

Page 34: Hyperdimensional Cognitive Computing: A New Approach to ......Hyperdimensional Cognitive Computing: A New Approach to Some Very Old Problems Simon D. Levy Washington & Lee University

Reduction(Holographic; Plate 2003)

Tuesday, April 14, 2009

Page 35: Hyperdimensional Cognitive Computing: A New Approach to ......Hyperdimensional Cognitive Computing: A New Approach to Some Very Old Problems Simon D. Levy Washington & Lee University

Reduction(Binary;

Kanerva 1994,Gayler 1998)

Tuesday, April 14, 2009

Page 36: Hyperdimensional Cognitive Computing: A New Approach to ......Hyperdimensional Cognitive Computing: A New Approach to Some Very Old Problems Simon D. Levy Washington & Lee University

Composition / Recursion

Tuesday, April 14, 2009

Page 37: Hyperdimensional Cognitive Computing: A New Approach to ......Hyperdimensional Cognitive Computing: A New Approach to Some Very Old Problems Simon D. Levy Washington & Lee University

Variables

X

john

Tuesday, April 14, 2009

Page 38: Hyperdimensional Cognitive Computing: A New Approach to ......Hyperdimensional Cognitive Computing: A New Approach to Some Very Old Problems Simon D. Levy Washington & Lee University

Scaling Up

• With many (> 10K) dimensions, get

• Astronomically large # of mutually orthogonal vectors (symbols)

• Surprising robustness to noise

Tuesday, April 14, 2009

Page 39: Hyperdimensional Cognitive Computing: A New Approach to ......Hyperdimensional Cognitive Computing: A New Approach to Some Very Old Problems Simon D. Levy Washington & Lee University

Pitfalls to Avoid

2. The Homunculus problem, a.k.a. Ghost in the Machine(Ryle 1949)

In cognitive modeling, the homunculus is the researcher: supervises learning, hand-builds representations, etc.

Tuesday, April 14, 2009

Page 40: Hyperdimensional Cognitive Computing: A New Approach to ......Hyperdimensional Cognitive Computing: A New Approach to Some Very Old Problems Simon D. Levy Washington & Lee University

Banishing the Homunculus

Tuesday, April 14, 2009

Page 41: Hyperdimensional Cognitive Computing: A New Approach to ......Hyperdimensional Cognitive Computing: A New Approach to Some Very Old Problems Simon D. Levy Washington & Lee University

Step I: Automatic Variable Substitution

• If A is a vector over {+1,-1}, then A*A = vector of 1’s (multiplicative identity)

• Supports substitution of anything for anything: everything (names, individuals, structures, propositions) can be a variable!

Tuesday, April 14, 2009

Page 42: Hyperdimensional Cognitive Computing: A New Approach to ......Hyperdimensional Cognitive Computing: A New Approach to Some Very Old Problems Simon D. Levy Washington & Lee University

“What is the Dollar of Mexico?” (Kanerva, to appear)

• Let X = <country>, Y = <currency>, A = <USA>, B = <Mexico>

• Then A = X*U + Y*D, B = X*M + Y*PD*A*B =

D*(X*U + Y*D) * (X*M + Y*P) =

(D*X*U + D*Y*D) * (X*M + Y*P) =

(D*X*U + Y) * (X*M + Y*P) =

D*X*U*X*M + D*X*U*Y*P + Y*X*M + Y*Y*P =

P + noise

Tuesday, April 14, 2009

Page 43: Hyperdimensional Cognitive Computing: A New Approach to ......Hyperdimensional Cognitive Computing: A New Approach to Some Very Old Problems Simon D. Levy Washington & Lee University

Learning Grammatical Constructions from a Single Example (Levy, to appear)

• Given

• Meaning: KISS(MARY, JOHN)

• Form: Mary kissed John

• Lexicon: KISS/kiss, MARY/Mary, ...

• What is the form for HIT(BILL, FRED) ?

Tuesday, April 14, 2009

Page 44: Hyperdimensional Cognitive Computing: A New Approach to ......Hyperdimensional Cognitive Computing: A New Approach to Some Very Old Problems Simon D. Levy Washington & Lee University

Learning Grammatical Constructions from a Single Example (Levy, to appear)

(ACTION*KISS + AGENT*MARY + PATIENT*JOHN) *

(P1*Mary + P2*kissed + P3*John) *

(KISS*kissed + MAY*Mary + JOHN*John + BILL*Bill + FRED*Fred + HIT*hit) *

(ACTION*HIT + AGENT*BILL + PATIENT*FRED) =

....

= (P1*Bill + P2*hit + P3*Fred) + noise

Tuesday, April 14, 2009

Page 45: Hyperdimensional Cognitive Computing: A New Approach to ......Hyperdimensional Cognitive Computing: A New Approach to Some Very Old Problems Simon D. Levy Washington & Lee University

Step II: Distributed “Lateral Inhibition”

• Analogical mapping as holistic graph isomorphsm (Gayler & Levy, in progress)

cf. Pelillo (1999)

A

B

C D

P

Q

R S

Tuesday, April 14, 2009

Page 46: Hyperdimensional Cognitive Computing: A New Approach to ......Hyperdimensional Cognitive Computing: A New Approach to Some Very Old Problems Simon D. Levy Washington & Lee University

A

B

C D

P

Q

R S

Possibilities x: A*P + A*Q + A*R + A*S + ... + D*S

Evidence w: A*B*P*Q + A*B*P*R +...+ B*C*Q*R + .. + C*D*R*S

X*W = A*Q + B*R + ... + A*P + ... + D*S

Tuesday, April 14, 2009

Page 47: Hyperdimensional Cognitive Computing: A New Approach to ......Hyperdimensional Cognitive Computing: A New Approach to Some Very Old Problems Simon D. Levy Washington & Lee University

c∧

c

w * cleanup /∑

xt

xt+1πt

Tuesday, April 14, 2009

Page 48: Hyperdimensional Cognitive Computing: A New Approach to ......Hyperdimensional Cognitive Computing: A New Approach to Some Very Old Problems Simon D. Levy Washington & Lee University

Tuesday, April 14, 2009

Page 49: Hyperdimensional Cognitive Computing: A New Approach to ......Hyperdimensional Cognitive Computing: A New Approach to Some Very Old Problems Simon D. Levy Washington & Lee University

Tuesday, April 14, 2009

Page 50: Hyperdimensional Cognitive Computing: A New Approach to ......Hyperdimensional Cognitive Computing: A New Approach to Some Very Old Problems Simon D. Levy Washington & Lee University

Future Work: Automatic Decomposition

MSC (Arathorn 2002)Tuesday, April 14, 2009

Page 51: Hyperdimensional Cognitive Computing: A New Approach to ......Hyperdimensional Cognitive Computing: A New Approach to Some Very Old Problems Simon D. Levy Washington & Lee University

MSC (Arathorn 2002)Tuesday, April 14, 2009