1 / 33

Saturday, December 13 Morning session

NIPS 2003 Workshop on Information Theory and Learning: The Bottleneck and Distortion Approach Organizers: Thomas Gedeon Naftali Tishby http://www.cs.huji.ac.il/~tishby/NIPS-Workshop. Saturday, December 13 Morning session

drea
Télécharger la présentation

Saturday, December 13 Morning session

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. NIPS 2003 Workshop onInformation Theory and Learning:The Bottleneck and Distortion ApproachOrganizers:Thomas Gedeon Naftali Tishbyhttp://www.cs.huji.ac.il/~tishby/NIPS-Workshop

  2. Saturday, December 13 Morning session • 7:30-8:20 N. TishbyIntroduction – A new look at Shannon’s Information Theory • 8:20-9:00 T. GedeonMathematical structure of Information Distortion methods • 9:00-9:30 D. Miller Deterministic annealing • 9:30-10:00 N. SlonimMultivariate Information Bottleneck • 10:00-10:30 B. MumeyOptimal Mutual Information Quantization is NP-complete Afternoon session • 16:00-16:20 G. Elidan The Information Bottleneck EM algorithm • 16:20-16:40 A. GlobersonSufficient Dimensionality Reduction with Side Information • 16:40-17:00 A. ParkerPhase Transitions in the Information Distortion • 17:00-17:20 A. Dimitrov Information Distortion as a model of sensory processing break • 17:40-18:00 J. SinkkonenIB-type clustering for continuous finite data • 18:00-18:20 G. ChechikGIB: Gaussian Information Bottleneck • 18:20-18:40 Y. CrammerIB and data-representation – Bregman to the rescue • 18:40-19:00 B. WestoverAchievable rates for Pattern Recognition

  3. Outline: The fundamental dilemma: Simplicity/Complexity versus Accuracy • Lessons from Statistical Physics for Machine Learning • Is there a “right level” of description? • A Variational Principle • Shannon’s source and channel coding – dual problems • [Mutual] Information as theFundamental Quantity • Information Bottleneck (IB) andEfficient Representations • Finite Data Issues • Algorithms, Applications and Extensions • Words, documents and meaning… • Understanding neural codes … • Quantization of side-information • Relevant linear dimension reduction: Gaussian IB • Using “irrelevant” side information • Multivariate-IB and Graphical Models • Sufficient Dimensionality Reduction (SDR) • The importance of being principled…

  4. Learning Theory:The fundamental dilemma… y=f(x) Tradeoff between accuracy and simplicity Y Good models should enable Prediction of new data… X

  5. Or in unsupervised learning … Y Cluster models? X

  6. Lessons from Statistical Physics Physical systems with many degrees of freedom: What is the right level of description? • “Microscopic Variables” – dynamical variables (positions, momenta,…) • “Variational Functions” – Thermodynamic potentials (Energy, Entropy, Free-energy,…) • Competition between order (energy) and disorder (entropy)… Emergent “relevant” description Order Parameters (magnetization,…)

  7. Statistical Models in Learning and AI Statistical models of complex systems: What is the right level of description? • “Microscopic Variables” – probability distributions (over all observed and hidden variables) • “Variational Functions” – Log-likelihood, Entropy,… ?(Multi-Information, “Free-energy”,… ) • Competition between accuracy and complexity Emerged “relevant” description Features (clusters, sufficient statistics,…)

  8. Limited data Bounded Computation The Fundamental Dilemma (of science): Model Complexity vs Prediction Accuracy Accuracy Survivability Possible Models/representations Complexity Efficiency

  9. Can we quantify it…? When there is a (relevant) prediction or distortion measure Accuracy small average distortion (good predictions) Complexity short description (high compression) A general tradeoff between distortion and compression: Shannon’s Information Theory

  10. decoder encoder receiver source channel Shannon’s Information Theory Sent messages Received messages symbols Source coding Channel coding Channel decoding Source decoding 0110101001110… Compression Error Correction Decompression Source Entropy Channel Capacity Rate vs Distortion Capacity vs Efficiency

  11. Compression and Mutual Information We need to index the max number of non-overlapping green blobs inside the blue blob: (mutual information!)

  12. Axioms for “(multi) information”(w I Nemenman) • M1: For two nodes, X and Y, the shared information,I[X;Y], is a continuous function of p(x,y). • M2: when one of the variables defines the other uniquely (1-1 mapping from X to Y) and both p(x)=p(y)=1/K , then I[X;Y] is a monotonic function of K. • M3: If X is partitioned into two subsets, X1 and X2, then the amount of shared information is additive in the following sense: I[X;Y]=I[(X1,X2);Y]+p(x1)I[X;Y|X1]+p(x2)I[X;Y|X2] • M4: Shared information is symmetric: I[X;Y]=I[Y;X]. • M5: Theorem:

  13. The Dual Variational problems of IT:Rate vs Distortion and Capacity vs Efficiency Q:What is the simplest representation – fewer bits(/sec) (Rate) for a given expected distortion (accuracy)? A: (RDT, Shannon 1960) solve: Q: What is the maximum number of bits(/sec) that can be sent reliably (prediction rate) for a given efficiency of the channel (power, cost)? A: (Capacity-power tradeoff, Shannon 1956) solve:

  14. Rate-Distortion theory • The tradeoff between expected representation size (Rate) and the expected distortion is expressed by the rate-distortion function: • By introducing a Lagrange multiplier, , we have a variational principle: • with:

  15. The Rate-Distortion function, R(D), is the optimal ratefor agiven distortionand is a convex function. R(D) Achievable region -bits/accuracy impossible D

  16. 0 The Capacity – Efficiency Tradeoff bits/erg Achievable Region C(E) Capacity at E E (Efficiency (power))

  17. Double matching of source to channel- eliminates the need for coding (?!) R(D) C(E) E D In the case of a Gaussian channel (w. power constraint) and square distortion, double matching means: And No coding!

  18. “There isa curious and provocative duality between the properties of a source with a distortion measure and those of a channel … if we consider channels in which there is a “cost” associated with the different input letters… The solution to this problem amounts, mathematically, to maximizing a mutual information under linear inequality constraint… which leads to a capacity-cost function C(E) for the channel… In a somewhat dual way, evaluating the rate-distortion function R(D) for source amounts, mathematically, to minimizing a mutual Information … again with a linear inequality constraint. This duality can be pursued further and is related to the duality between past and future and the notions of control and knowledge. Thus we may have knowledge of the past but cannot control it; we may control the future but have no knowledge of it.” C. E. Shannon (1959)

  19. Bottlenecks and Neural Nets • Auto association: forcing compact representations • provide a “relevant code” of w.r.t. Input Output Sample 1 Sample 2 Past Future

  20. Bottlenecks in Nature… (w E Schniedman, Rob de Ruyter, W Bialek, N Brenner)

  21. The idea: find a compressed variable that enables short encoding of ( small ) while preserving as much as possible the information on the relevant signal ( )

  22. A Variational Principle We want a short representation of X that keeps the information about another variable, Y, if possible.

  23. Self Consistent Equations • Marginal: • Markov condition: • Bayes’ rule:

  24. The emerged effective distortion measure: • Regular if is absolutely continuous w.r.t. • Small if predicts y as well as x:

  25. I-Projections on a set of distributions (Csiszar 75,84) • The I-projection of a distribution q(x) on a convex set of distributions L:

  26. The Blahut-Arimoto Algorithm

  27. The Information BottleneckAlgorithm “free energy”

  28. The IB emergenteffective distortion measure: The IB algorithm

  29. Key Theorems IB Coding Theorem(Wyner 1975, Bachrach, Navot, Tishby 2003): The IB variational problem provides a tight convex lower bound on the expected size of the achievable representations of X, that maintain at least of mutual information on the variable Y. The bound can be achieved by: . Formally equivalent to “Channel coding with Side-Information” (introduced in a very different context by Wyner, 75). The same bound is true for the dual channel coding problem and the optimal representation constitutes a “perfectly matched source-channel” and requires “no further coding”. IB Algorithm Convergence(Tishby 99, Friedman, Slonim,Tishby 01): The GeneralizedArimoto-Blauht IB algorithm converges to a local minima of the IB functional, (including its generalized multivariate case).

  30. Summary • Shannon’s Information Theory suggests a compelling framework for quantifying the “Complexity-Accuracy” dilemma, by unifying source and channel coding as a Min Max of mutual information. • The IB method turns this unified coding principle into algorithms for extracting “informative” relevant structures from empirical joint probability distributions P(X1,X2)… • The Multivariate IB extends this framework to extract “informative” structures from multivariate distributions, P(X1,…,Xn), via Min Max I(X1,…,Xn),and Graphical models • IB can be further extended for Sufficient Dimensionality Reduction a method for finding informative continuous low dimensional representations via Min Max I(X1,X2)

  31. Many thanks to: Fernando Pereira William Bialek Nir Friedman Noam Slonim Gal Chechik Amir Globerson Ran Bachrach Amir Navot Ilya Nemenman

More Related