Presentation is loading. Please wait.

Presentation is loading. Please wait.

Statistical Relational Learning Pedro Domingos Dept. of Computer Science & Eng. University of Washington.

Similar presentations


Presentation on theme: "Statistical Relational Learning Pedro Domingos Dept. of Computer Science & Eng. University of Washington."— Presentation transcript:

1 Statistical Relational Learning Pedro Domingos Dept. of Computer Science & Eng. University of Washington

2 Overview Motivation Background Representation Inference Learning Software Applications Discussion

3 Motivation Most learners assume i.i.d. data (independent and identically distributed) One type of object Objects have no relation to each other Real applications: dependent, variously distributed data Multiple types of objects Relations between objects

4 Examples Web search Information extraction Natural language processing Perception Medical diagnosis Computational biology Social networks Ubiquitous computing Etc.

5 Costs and Benefits of SRL Benefits Better predictive accuracy Better understanding of domains Growth path for machine learning Costs Learning is much harder Inference becomes a crucial issue Greater complexity for user

6 Goal and Progress Goal: Learn from non-i.i.d. data as easily as from i.i.d. data Progress to date Burgeoning research area We’re “close enough” to goal Open-source software available Lots of research questions (old and new)

7 Plan We have the elements: Probability for handling uncertainty Logic for representing types, relations, and complex dependencies between them Learning and inference algorithms for each Figure out how to put them together Tremendous leverage on a wide range of applications

8 Disclaimers Not a complete survey of statistical relational learning Or of foundational areas Focus is practical, not theoretical Assumes basic background in logic, probability and statistics, etc. Please ask questions Tutorial and examples available at alchemy.cs.washington.edu

9 Overview Motivation Background Representation Inference Learning Software Applications Discussion

10 Markov Networks Undirected graphical models Cancer CoughAsthma Smoking Potential functions defined over cliques SmokingCancer Ф(S,C) False 4.5 FalseTrue 4.5 TrueFalse 2.7 True 4.5

11 Markov Networks Undirected graphical models Log-linear model: Weight of Feature iFeature i Cancer CoughAsthma Smoking

12 First-Order Logic Constants, variables, functions, predicates E.g.: Anna, X, mother_of(X), friends(X, Y) Grounding: Replace all variables by constants E.g.: friends (Anna, Bob) World (model, interpretation): Assignment of truth values to all ground predicates

13 Overview Motivation Background Representation Inference Learning Software Applications Discussion

14 Representations RepresentationLogical Language Probabilistic Language Knowledge-based model construction Horn clausesBayes nets Stochastic logic programs Horn clausesPCFGs Probabilistic relational models Frame systemsBayes nets Relational Markov networks SQL queriesMarkov nets Bayesian logicFirst-orderBayes nets Markov logicFirst-order logicMarkov nets

15 Markov Logic Most developed approach to date Many other approaches can be viewed as special cases Used in rest of this tutorial

16 Markov Logic: Intuition A logical KB is a set of hard constraints on the set of possible worlds Let’s make them soft constraints: When a world violates a formula, It becomes less probable, not impossible Give each formula a weight (Higher weight  Stronger constraint)

17 Markov Logic: Definition A Markov Logic Network (MLN) is a set of pairs (F, w) where F is a formula in first-order logic w is a real number Together with a set of constants, it defines a Markov network with One node for each grounding of each predicate in the MLN One feature for each grounding of each formula F in the MLN, with the corresponding weight w

18 Example: Friends & Smokers

19

20

21

22 Two constants: Anna (A) and Bob (B)

23 Example: Friends & Smokers Cancer(A) Smokes(A)Smokes(B) Cancer(B) Two constants: Anna (A) and Bob (B)

24 Example: Friends & Smokers Cancer(A) Smokes(A)Friends(A,A) Friends(B,A) Smokes(B) Friends(A,B) Cancer(B) Friends(B,B) Two constants: Anna (A) and Bob (B)

25 Example: Friends & Smokers Cancer(A) Smokes(A)Friends(A,A) Friends(B,A) Smokes(B) Friends(A,B) Cancer(B) Friends(B,B) Two constants: Anna (A) and Bob (B)

26 Example: Friends & Smokers Cancer(A) Smokes(A)Friends(A,A) Friends(B,A) Smokes(B) Friends(A,B) Cancer(B) Friends(B,B) Two constants: Anna (A) and Bob (B)

27 Markov Logic Networks MLN is template for ground Markov nets Probability of a world x : Typed variables and constants greatly reduce size of ground Markov net Functions, existential quantifiers, etc. Infinite and continuous domains Weight of formula iNo. of true groundings of formula i in x

28 Relation to Statistical Models Special cases: Markov networks Markov random fields Bayesian networks Log-linear models Exponential models Max. entropy models Gibbs distributions Boltzmann machines Logistic regression Hidden Markov models Conditional random fields Obtained by making predicates zero-arity Markov logic allows objects to be interdependent (non-i.i.d.) Easy to compose models

29 Relation to First-Order Logic Infinite weights  First-order logic Satisfiable KB, positive weights  Satisfying assignments = Modes of distribution Markov logic allows contradictions between formulas

30 Overview Motivation Background Representation Inference Learning Software Applications Discussion

31 Inference Goal: Compute marginal probabilities of nodes (or formulas) given evidence Approaches: Markov chain Monte Carlo Belief propagation Variational approximations Etc. Other inference tasks: Compute most likely state of world Compute actions that maximize utility Etc.

32 Lifted Inference We can do inference in first-order logic without grounding the KB (e.g.: resolution) Let’s do the same for inference in MLNs Group atoms and clauses into “indistinguishable” sets Do inference over those First approach: Lifted variable elimination (not practical) Here: Lifted belief propagation

33 Belief Propagation Nodes (x) Features (f)

34 Lifted Belief Propagation Nodes (x) Features (f)

35 Lifted Belief Propagation Nodes (x) Features (f)

36 Lifted Belief Propagation   ,  : Functions of edge counts Nodes (x) Features (f)

37 Lifted Belief Propagation Form lifted network composed of supernodes and superfeatures Supernode: Set of ground atoms that all send and receive same messages throughout BP Superfeature: Set of ground clauses that all send and receive same messages throughout BP Run belief propagation on lifted network Guaranteed to produce same results as ground BP Time and memory savings can be huge

38 Forming the Lifted Network 1. Form initial supernodes One per predicate and truth value (true, false, unknown) 2. Form superfeatures by doing joins of their supernodes 3. Form supernodes by projecting superfeatures down to their predicates Supernode = Groundings of a predicate with same number of projections from each superfeature 4. Repeat until convergence

39 Theorem There exists a unique minimal lifted network The lifted network construction algo. finds it BP on lifted network gives same result as on ground network

40 Representing Supernodes And Superfeatures List of tuples: Simple but inefficient Resolution-like: Use equality and inequality Form clusters of atoms and clauses

41 Overview Motivation Background Representation Inference Learning Software Applications Discussion

42 Learning Data is a relational database Closed world assumption (if not: EM) Learning parameters (weights) Generatively Discriminatively Learning structure (formulas)

43 Generative Weight Learning Maximize likelihood Use gradient ascent or L-BFGS No local maxima Requires inference at each step (slow!) No. of true groundings of clause i in data Expected no. true groundings according to model

44 Pseudo-Likelihood Likelihood of each variable given its neighbors in the data Does not require inference at each step Consistent estimator Widely used in vision, spatial statistics, etc. But PL parameters may not work well for long inference chains

45 Discriminative Weight Learning Maximize conditional likelihood of query ( y ) given evidence ( x ) Optimization: Scaled conjugate gradient, diagonal Newton, etc. No. of true groundings of clause i in data Expected no. true groundings according to model

46 Structure Learning Generalizes feature induction in Markov nets Any inductive logic programming approach can be used, but... Goal is to induce any clauses, not just Horn Evaluation function should be likelihood Requires learning weights for each candidate Turns out not to be bottleneck Bottleneck is counting clause groundings Solution: Subsampling

47 Structure Learning Initial state: Unit clauses or hand-coded KB Operators: Add/remove literal, flip sign Evaluation function: Pseudo-likelihood + Structure prior Search: Beam, shortest-first, bottom-up, stochastic, etc.

48 Overview Motivation Background Representation Inference Learning Software Applications Discussion

49 Alchemy Open-source software including: Full first-order logic syntax Generative & discriminative weight learning Structure learning Inference (marginals and most prob. states) Programming language features www.cs.washington.edu/ai/alchemy

50 AlchemyPrologBUGS Represent- ation F.O. Logic + Markov nets Horn clauses Bayes nets InferenceLifted BP, etc.Theorem proving MCMC LearningParameters & structure NoParams. UncertaintyYesNoYes RelationalYes No

51 Overview Motivation Background Representation Inference Learning Software Applications Discussion

52 Applications Information extraction Entity resolution Link prediction Collective classification Web mining Natural language processing Computational biology Social network analysis Robot mapping Activity recognition Probabilistic Cyc CALO Etc.

53 Information Extraction Parag Singla and Pedro Domingos, “Memory-Efficient Inference in Relational Domains” (AAAI-06). Singla, P., & Domingos, P. (2006). Memory-efficent inference in relatonal domains. In Proceedings of the Twenty-First National Conference on Artificial Intelligence (pp. 500-505). Boston, MA: AAAI Press. H. Poon & P. Domingos, Sound and Efficient Inference with Probabilistic and Deterministic Dependencies”, in Proc. AAAI-06, Boston, MA, 2006. P. Hoifung (2006). Efficent inference. In Proceedings of the Twenty-First National Conference on Artificial Intelligence.

54 Segmentation Parag Singla and Pedro Domingos, “Memory-Efficient Inference in Relational Domains” (AAAI-06). Singla, P., & Domingos, P. (2006). Memory-efficent inference in relatonal domains. In Proceedings of the Twenty-First National Conference on Artificial Intelligence (pp. 500-505). Boston, MA: AAAI Press. H. Poon & P. Domingos, Sound and Efficient Inference with Probabilistic and Deterministic Dependencies”, in Proc. AAAI-06, Boston, MA, 2006. P. Hoifung (2006). Efficent inference. In Proceedings of the Twenty-First National Conference on Artificial Intelligence. Author Title Venue

55 Entity Resolution Parag Singla and Pedro Domingos, “Memory-Efficient Inference in Relational Domains” (AAAI-06). Singla, P., & Domingos, P. (2006). Memory-efficent inference in relatonal domains. In Proceedings of the Twenty-First National Conference on Artificial Intelligence (pp. 500-505). Boston, MA: AAAI Press. H. Poon & P. Domingos, Sound and Efficient Inference with Probabilistic and Deterministic Dependencies”, in Proc. AAAI-06, Boston, MA, 2006. P. Hoifung (2006). Efficent inference. In Proceedings of the Twenty-First National Conference on Artificial Intelligence.

56 Entity Resolution Parag Singla and Pedro Domingos, “Memory-Efficient Inference in Relational Domains” (AAAI-06). Singla, P., & Domingos, P. (2006). Memory-efficent inference in relatonal domains. In Proceedings of the Twenty-First National Conference on Artificial Intelligence (pp. 500-505). Boston, MA: AAAI Press. H. Poon & P. Domingos, Sound and Efficient Inference with Probabilistic and Deterministic Dependencies”, in Proc. AAAI-06, Boston, MA, 2006. P. Hoifung (2006). Efficent inference. In Proceedings of the Twenty-First National Conference on Artificial Intelligence.

57 State of the Art Segmentation HMM (or CRF) to assign each token to a field Entity resolution Logistic regression to predict same field/citation Transitive closure Alchemy implementation: Seven formulas

58 Types and Predicates token = {Parag, Singla, and, Pedro,...} field = {Author, Title, Venue} citation = {C1, C2,...} position = {0, 1, 2,...} Token(token, position, citation) InField(position, field, citation) SameField(field, citation, citation) SameCit(citation, citation)

59 Types and Predicates token = {Parag, Singla, and, Pedro,...} field = {Author, Title, Venue,...} citation = {C1, C2,...} position = {0, 1, 2,...} Token(token, position, citation) InField(position, field, citation) SameField(field, citation, citation) SameCit(citation, citation) Optional

60 Types and Predicates Evidence token = {Parag, Singla, and, Pedro,...} field = {Author, Title, Venue} citation = {C1, C2,...} position = {0, 1, 2,...} Token(token, position, citation) InField(position, field, citation) SameField(field, citation, citation) SameCit(citation, citation)

61 token = {Parag, Singla, and, Pedro,...} field = {Author, Title, Venue} citation = {C1, C2,...} position = {0, 1, 2,...} Token(token, position, citation) InField(position, field, citation) SameField(field, citation, citation) SameCit(citation, citation) Types and Predicates Query

62 Token(+t,i,c) => InField(i,+f,c) InField(i,+f,c) InField(i+1,+f,c) f != f’ => (!InField(i,+f,c) v !InField(i,+f’,c)) Token(+t,i,c) ^ InField(i,+f,c) ^ Token(+t,i’,c’) ^ InField(i’,+f,c’) => SameField(+f,c,c’) SameField(+f,c,c’) SameCit(c,c’) SameField(f,c,c’) ^ SameField(f,c’,c”) => SameField(f,c,c”) SameCit(c,c’) ^ SameCit(c’,c”) => SameCit(c,c”) Formulas

63 Token(+t,i,c) => InField(i,+f,c) InField(i,+f,c) InField(i+1,+f,c) f != f’ => (!InField(i,+f,c) v !InField(i,+f’,c)) Token(+t,i,c) ^ InField(i,+f,c) ^ Token(+t,i’,c’) ^ InField(i’,+f,c’) => SameField(+f,c,c’) SameField(+f,c,c’) SameCit(c,c’) SameField(f,c,c’) ^ SameField(f,c’,c”) => SameField(f,c,c”) SameCit(c,c’) ^ SameCit(c’,c”) => SameCit(c,c”)

64 Formulas Token(+t,i,c) => InField(i,+f,c) InField(i,+f,c) InField(i+1,+f,c) f != f’ => (!InField(i,+f,c) v !InField(i,+f’,c)) Token(+t,i,c) ^ InField(i,+f,c) ^ Token(+t,i’,c’) ^ InField(i’,+f,c’) => SameField(+f,c,c’) SameField(+f,c,c’) SameCit(c,c’) SameField(f,c,c’) ^ SameField(f,c’,c”) => SameField(f,c,c”) SameCit(c,c’) ^ SameCit(c’,c”) => SameCit(c,c”)

65 Formulas Token(+t,i,c) => InField(i,+f,c) InField(i,+f,c) InField(i+1,+f,c) f != f’ => (!InField(i,+f,c) v !InField(i,+f’,c)) Token(+t,i,c) ^ InField(i,+f,c) ^ Token(+t,i’,c’) ^ InField(i’,+f,c’) => SameField(+f,c,c’) SameField(+f,c,c’) SameCit(c,c’) SameField(f,c,c’) ^ SameField(f,c’,c”) => SameField(f,c,c”) SameCit(c,c’) ^ SameCit(c’,c”) => SameCit(c,c”)

66 Token(+t,i,c) => InField(i,+f,c) InField(i,+f,c) InField(i+1,+f,c) f != f’ => (!InField(i,+f,c) v !InField(i,+f’,c)) Token(+t,i,c) ^ InField(i,+f,c) ^ Token(+t,i’,c’) ^ InField(i’,+f,c’) => SameField(+f,c,c’) SameField(+f,c,c’) SameCit(c,c’) SameField(f,c,c’) ^ SameField(f,c’,c”) => SameField(f,c,c”) SameCit(c,c’) ^ SameCit(c’,c”) => SameCit(c,c”) Formulas

67 Token(+t,i,c) => InField(i,+f,c) InField(i,+f,c) InField(i+1,+f,c) f != f’ => (!InField(i,+f,c) v !InField(i,+f’,c)) Token(+t,i,c) ^ InField(i,+f,c) ^ Token(+t,i’,c’) ^ InField(i’,+f,c’) => SameField(+f,c,c’) SameField(+f,c,c’) SameCit(c,c’) SameField(f,c,c’) ^ SameField(f,c’,c”) => SameField(f,c,c”) SameCit(c,c’) ^ SameCit(c’,c”) => SameCit(c,c”) Formulas

68 Token(+t,i,c) => InField(i,+f,c) InField(i,+f,c) InField(i+1,+f,c) f != f’ => (!InField(i,+f,c) v !InField(i,+f’,c)) Token(+t,i,c) ^ InField(i,+f,c) ^ Token(+t,i’,c’) ^ InField(i’,+f,c’) => SameField(+f,c,c’) SameField(+f,c,c’) SameCit(c,c’) SameField(f,c,c’) ^ SameField(f,c’,c”) => SameField(f,c,c”) SameCit(c,c’) ^ SameCit(c’,c”) => SameCit(c,c”)

69 Formulas Token(+t,i,c) => InField(i,+f,c) InField(i,+f,c) ^ !Token(“.”,i,c) InField(i+1,+f,c) f != f’ => (!InField(i,+f,c) v !InField(i,+f’,c)) Token(+t,i,c) ^ InField(i,+f,c) ^ Token(+t,i’,c’) ^ InField(i’,+f,c’) => SameField(+f,c,c’) SameField(+f,c,c’) SameCit(c,c’) SameField(f,c,c’) ^ SameField(f,c’,c”) => SameField(f,c,c”) SameCit(c,c’) ^ SameCit(c’,c”) => SameCit(c,c”)

70 Results: Segmentation on Cora

71 Results: Matching Venues on Cora

72 Overview Motivation Background Representation Inference Learning Software Applications Discussion

73 Summary The real world is complex and uncertain First-order logic handles complexity Probability handles uncertainty Statistical relational learning combines the two Markov logic: Most advanced approach to date Alchemy (alchemy.cs.washington.edu): Complete suite of state-of-the-art algorithms Many challenging applications now within reach We’re at an inflection point in what we can do

74


Download ppt "Statistical Relational Learning Pedro Domingos Dept. of Computer Science & Eng. University of Washington."

Similar presentations


Ads by Google