Readings: K&F: 11.3, 11.5 Yedidia et al. paper from the class website

Slides:



Advertisements
Similar presentations
1 Undirected Graphical Models Graphical Models – Carlos Guestrin Carnegie Mellon University October 29 th, 2008 Readings: K&F: 4.1, 4.2, 4.3, 4.4,
Advertisements

Discrete Optimization Lecture 4 – Part 3 M. Pawan Kumar Slides available online
Exact Inference. Inference Basic task for inference: – Compute a posterior distribution for some query variables given some observed evidence – Sum out.
Variational Inference Amr Ahmed Nov. 6 th Outline Approximate Inference Variational inference formulation – Mean Field Examples – Structured VI.
Variational Methods for Graphical Models Micheal I. Jordan Zoubin Ghahramani Tommi S. Jaakkola Lawrence K. Saul Presented by: Afsaneh Shirazi.
Constrained Approximate Maximum Entropy Learning (CAMEL) Varun Ganapathi, David Vickrey, John Duchi, Daphne Koller Stanford University TexPoint fonts used.
Section 3: Appendix BP as an Optimization Algorithm 1.
Exact Inference in Bayes Nets
Parameter Learning in MN. Outline CRF Learning CRF for 2-d image segmentation IPF parameter sharing revisited.
Junction Trees And Belief Propagation. Junction Trees: Motivation What if we want to compute all marginals, not just one? Doing variable elimination for.
Loopy Belief Propagation a summary. What is inference? Given: –Observabled variables Y –Hidden variables X –Some model of P(X,Y) We want to make some.
Undirected Probabilistic Graphical Models (Markov Nets) (Slides from Sam Roweis)
An Introduction to Variational Methods for Graphical Models.
Convergent Message-Passing Algorithms for Inference over General Graphs with Convex Free Energies Tamir Hazan, Amnon Shashua School of Computer Science.
Introduction to Belief Propagation and its Generalizations. Max Welling Donald Bren School of Information and Computer and Science University of California.
Clique Trees Amr Ahmed October 23, Outline Clique Trees Representation Factorization Inference Relation with VE.
Belief Propagation by Jakob Metzler. Outline Motivation Pearl’s BP Algorithm Turbo Codes Generalized Belief Propagation Free Energies.
Overview of Inference Algorithms for Bayesian Networks Wei Sun, PhD Assistant Research Professor SEOR Dept. & C4I Center George Mason University, 2009.
Markov Networks.
Belief Propagation on Markov Random Fields Aggeliki Tsoli.
CS774. Markov Random Field : Theory and Application Lecture 04 Kyomin Jung KAIST Sep
Junction Trees: Motivation Standard algorithms (e.g., variable elimination) are inefficient if the undirected graph underlying the Bayes Net contains cycles.
Junction tree Algorithm :Probabilistic Graphical Models Recitation: 10/04/07 Ramesh Nallapati.
GS 540 week 6. HMM basics Given a sequence, and state parameters: – Each possible path through the states has a certain probability of emitting the sequence.
Parameter Learning in Markov Nets Dhruv Batra, Recitation 11/13/2008.
Global Approximate Inference Eran Segal Weizmann Institute.
Belief Propagation, Junction Trees, and Factor Graphs
Some Surprises in the Theory of Generalized Belief Propagation Jonathan Yedidia Mitsubishi Electric Research Labs (MERL) Collaborators: Bill Freeman (MIT)
Adaptive CSMA under the SINR Model: Fast convergence using the Bethe Approximation Krishna Jagannathan IIT Madras (Joint work with) Peruru Subrahmanya.
Undirected Models: Markov Networks David Page, Fall 2009 CS 731: Advanced Methods in Artificial Intelligence, with Biomedical Applications.
1 Structured Region Graphs: Morphing EP into GBP Max Welling Tom Minka Yee Whye Teh.
Belief Propagation. What is Belief Propagation (BP)? BP is a specific instance of a general class of methods that exist for approximate inference in Bayes.
1 Variable Elimination Graphical Models – Carlos Guestrin Carnegie Mellon University October 11 th, 2006 Readings: K&F: 8.1, 8.2, 8.3,
Daphne Koller Message Passing Belief Propagation Algorithm Probabilistic Graphical Models Inference.
1 Mean Field and Variational Methods finishing off Graphical Models – Carlos Guestrin Carnegie Mellon University November 5 th, 2008 Readings: K&F:
Belief Propagation and its Generalizations Shane Oldenburger.
1 Parameter Learning 2 Structure Learning 1: The good Graphical Models – Carlos Guestrin Carnegie Mellon University September 27 th, 2006 Readings:
1 Use graphs and not pure logic Variables represented by nodes and dependencies by edges. Common in our language: “threads of thoughts”, “lines of reasoning”,
Exact Inference in Bayes Nets. Notation U: set of nodes in a graph X i : random variable associated with node i π i : parents of node i Joint probability:
1 Param. Learning (MLE) Structure Learning The Good Graphical Models – Carlos Guestrin Carnegie Mellon University October 1 st, 2008 Readings: K&F:
1 Learning P-maps Param. Learning Graphical Models – Carlos Guestrin Carnegie Mellon University September 24 th, 2008 Readings: K&F: 3.3, 3.4, 16.1,
1 BN Semantics 2 – Representation Theorem The revenge of d-separation Graphical Models – Carlos Guestrin Carnegie Mellon University September 17.
1 Relational Factor Graphs Lin Liao Joint work with Dieter Fox.
1 Structure Learning (The Good), The Bad, The Ugly Inference Graphical Models – Carlos Guestrin Carnegie Mellon University October 13 th, 2008 Readings:
1 Variable Elimination Graphical Models – Carlos Guestrin Carnegie Mellon University October 15 th, 2008 Readings: K&F: 8.1, 8.2, 8.3,
1 BN Semantics 1 Graphical Models – Carlos Guestrin Carnegie Mellon University September 15 th, 2006 Readings: K&F: 3.1, 3.2, 3.3.
Markov Networks.
Readings: K&F: 15.1, 15.2, 15.3, 15.4, 15.5 K&F: 7 (overview of inference) K&F: 8.1, 8.2 (Variable Elimination) Structure Learning in BNs 3: (the good,
Lecture 5 Unsupervised Learning in fully Observed Directed and Undirected Graphical Models.
Clique Tree & Independence
Graduate School of Information Sciences, Tohoku University
Physical Fluctuomatics 7th~10th Belief propagation
Expectation-Maximization & Belief Propagation
Variable Elimination 2 Clique Trees
Probabilistic image processing and Bayesian network
Parameter Learning 2 Structure Learning 1: The good
Readings: K&F: 5.1, 5.2, 5.3, 5.4, 5.5, 5.6, 5.7 Markov networks, Factor graphs, and an unified view Start approximate inference If we are lucky… Graphical.
Approximate Inference by Sampling
Unifying Variational and GBP Learning Parameters of MNs EM for BNs
Junction Trees 3 Undirected Graphical Models
Readings: K&F: 11.3, 11.5 Yedidia et al. paper from the class website
Markov Networks.
Variable Elimination Graphical Models – Carlos Guestrin
BP in Practice Message Passing Inference Probabilistic Graphical
Mean Field and Variational Methods Loopy Belief Propagation
Generalized Belief Propagation
Presentation transcript:

Readings: K&F: 11.3, 11.5 Yedidia et al. paper from the class website Chapter 9 - Jordan Loopy Belief Propagation Generalized Belief Propagation Unifying Variational and GBP Learning Parameters of MNs Graphical Models – 10708 Carlos Guestrin Carnegie Mellon University November 10th, 2006

More details on Loopy BP Difficulty SAT Grade Happy Job Coherence Letter Intelligence Numerical problem: messages < 1 get multiplied together as we go around the loops numbers can go to zero normalize messages to one: Zi!j doesn’t depend on Xj, so doesn’t change the answer Computing node “beliefs” (estimates of probs.): 10-708 – Carlos Guestrin 2006

An example of running loopy BP 10-708 – Carlos Guestrin 2006

(Non-)Convergence of Loopy BP Loopy BP can oscillate!!! oscillations can small oscillations can be really bad! Typically, if factors are closer to uniform, loopy does well (converges) if factors are closer to deterministic, loopy doesn’t behave well One approach to help: damping messages new message is average of old message and new one: often better convergence but, when damping is required to get convergence, result often bad 10-708 – Carlos Guestrin 2006 graphs from Murphy et al. ’99

Loopy BP in Factor graphs What if we don’t have pairwise Markov nets? Transform to a pairwise MN Use Loopy BP on a factor graph Message example: from node to factor: from factor to node: A B C D E ABC ABD BDE CDE 10-708 – Carlos Guestrin 2006

Loopy BP in Factor graphs From node i to factor j: F(i) factors whose scope includes Xi From factor j to node i: Scope[j] = Y[{Xi} A B C D E ABC ABD BDE CDE 10-708 – Carlos Guestrin 2006

What you need to know about loopy BP Application of belief propagation in loopy graphs Doesn’t always converge damping can help good message schedules can help (see book) If converges, often to incorrect, but useful results Generalizes from pairwise Markov networks by using factor graphs 10-708 – Carlos Guestrin 2006

Announcements Monday’s special recitation Pradeep Ravikumar on exciting new approximate inference algorithms 10-708 – Carlos Guestrin 2006

Loopy BP v. Clique trees: Two ends of a spectrum DIG GJSL HGJ CD GSI Difficulty SAT Grade Happy Job Coherence Letter Intelligence 10-708 – Carlos Guestrin 2006

Generalize cluster graph Generalized cluster graph: For set of factors F Undirected graph Each node i associated with a cluster Ci Family preserving: for each factor fj 2 F, 9 node i such that scope[fi]µ Ci Each edge i – j is associated with a set of variables Sij µ Ci Å Cj 10-708 – Carlos Guestrin 2006

Running intersection property (Generalized) Running intersection property (RIP) Cluster graph satisfies RIP if whenever X2 Ci and X2 Cj then 9 one and only one path from Ci to Cj where X2Suv for every edge (u,v) in the path 10-708 – Carlos Guestrin 2006

Examples of cluster graphs 10-708 – Carlos Guestrin 2006

Two cluster graph satisfying RIP with different edge sets 10-708 – Carlos Guestrin 2006

Generalized BP on cluster graphs satisfying RIP Initialization: Assign each factor  to a clique (), Scope[]µC() Initialize cliques: Initialize messages: While not converged, send messages: Belief: 10-708 – Carlos Guestrin 2006

Cluster graph for Loopy BP Difficulty SAT Grade Happy Job Coherence Letter Intelligence 10-708 – Carlos Guestrin 2006

What if the cluster graph doesn’t satisfy RIP 10-708 – Carlos Guestrin 2006

Region graphs to the rescue Can address generalized cluster graphs that don’t satisfy RIP using region graphs: Yedidia et al. from class website Example in your homework!  Hint – From Yedidia et al.: Section 7 – defines region graphs Section 9 – message passing on region graphs Section 10 – An example that will help you a lot!!!  10-708 – Carlos Guestrin 2006

Revisiting Mean-Fields Choice of Q: Optimization problem: 10-708 – Carlos Guestrin 2006

Interpretation of energy functional Exact if P=Q: View problem as an approximation of entropy term: 10-708 – Carlos Guestrin 2006

Entropy of a tree distribution Difficulty SAT Grade Happy Job Coherence Letter Intelligence Entropy term: Joint distribution: Decomposing entropy term: More generally: di number neighbors of Xi 10-708 – Carlos Guestrin 2006

Loopy BP & Bethe approximation Difficulty SAT Grade Happy Job Coherence Letter Intelligence Energy functional: Bethe approximation of Free Energy: use entropy for trees, but loopy graphs: Theorem: If Loopy BP converges, resulting ij & i are stationary point (usually local maxima) of Bethe Free energy! 10-708 – Carlos Guestrin 2006

GBP & Kikuchi approximation DIG GJSL HGJ CD GSI Exact Free energy: Junction Tree Bethe Free energy: Kikuchi approximation: Generalized cluster graph spectrum from Bethe to exact entropy terms weighted by counting numbers see Yedidia et al. Theorem: If GBP converges, resulting Ci are stationary point (usually local maxima) of Kikuchi Free energy! Difficulty SAT Grade Happy Job Coherence Letter Intelligence 10-708 – Carlos Guestrin 2006

What you need to know about GBP Spectrum between Loopy BP & Junction Trees: More computation, but typically better answers If satisfies RIP, equations are very simple General setting, slightly trickier equations, but not hard Relates to variational methods: Corresponds to local optima of approximate version of energy functional 10-708 – Carlos Guestrin 2006

Learning Parameters of a BN D S G H J C L I Log likelihood decomposes: Learn each CPT independently Use counts 10-708 – Carlos Guestrin 2006

Log Likelihood for MN Log likelihood of the data: Difficulty SAT Grade Happy Job Coherence Letter Intelligence 10-708 – Carlos Guestrin 2006

Log Likelihood doesn’t decompose for MNs A convex problem Can find global optimum!! Term log Z doesn’t decompose!! Difficulty SAT Grade Happy Job Coherence Letter Intelligence 10-708 – Carlos Guestrin 2006

Derivative of Log Likelihood for MNs Difficulty SAT Grade Happy Job Coherence Letter Intelligence 10-708 – Carlos Guestrin 2006

Derivative of Log Likelihood for MNs Difficulty SAT Grade Happy Job Coherence Letter Intelligence Derivative: Setting derivative to zero Can optimize using gradient ascent Let’s look at a simpler solution 10-708 – Carlos Guestrin 2006

Iterative Proportional Fitting (IPF) Difficulty SAT Grade Happy Job Coherence Letter Intelligence Setting derivative to zero: Fixed point equation: Iterate and converge to optimal parameters Each iteration, must compute: 10-708 – Carlos Guestrin 2006

What you need to know about learning MN parameters? BN parameter learning easy MN parameter learning doesn’t decompose! Learning requires inference! Apply gradient ascent or IPF iterations to obtain optimal parameters 10-708 – Carlos Guestrin 2006