1. Markov Logic Pedro Domingos Dept. of Computer Science & Eng. University of Washington

2. Desiderata A language for cognitive modeling should: Handle uncertainty Noise Incomplete information Ambiguity Handle complexity Many objects Relations among them IsA and IsPart hierarchies Etc.

3. Solution Probability handles uncertainty Logic handles complexity What is the simplest way to combine the two?

4. Markov Logic Assign weights to logical formulas Treat formulas as templates for features of Markov networks

5. Overview Representation Inference Learning Applications

6. Propositional Logic Atoms: Symbols representing propositions Logical connectives: ¬, Λ, V, etc. Knowledge base: Set of formulas World: Truth assignment to all atoms Every KB can be converted to CNF CNF: Conjunction of clauses Clause: Disjunction of literals Literal: Atom or its negation Entailment: Does KB entail query?

7. First-Order Logic Atom: Predicate(Variables,Constants) E.g.: Ground atom: All arguments are constants Quantifiers: This tutorial: Finite, Herbrand interpretations

8. 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

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

10. Probabilistic Knowledge Bases PKB = Set of formulas and their probabilities + Consistency + Maximum entropy = Set of formulas and their weights = Set of formulas and their potentials (1 if formula true, if formula false)

11. Markov Logic 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 An MLN 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

12. 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 all predicates zero-arity Markov logic allows objects to be interdependent (non-i.i.d.) Markov logic facilitates composition

13. 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

14. Example

19. Overview Representation Inference Learning Applications

20. Theorem Proving TP(KB, Query) KB Q ← KB U {¬ Query} return ¬SAT(CNF(KB Q ))

21. Satisfiability (DPLL) SAT(CNF) if CNF is empty return True if CNF contains empty clause return False choose an atom A return SAT(CNF(A)) V SAT(CNF(¬A))

22. First-Order Theorem Proving Propositionalization 1. Form all possible ground atoms 2. Apply propositional theorem prover Lifted Inference: Resolution Resolve pairs of clauses until empty clause derived Unify literals by substitution, e.g.: unifies and

23. Probabilistic Theorem Proving Given Probabilistic knowledge base K Query formula Q Output P(Q|K)

24. Weighted Model Counting ModelCount(CNF) = # worlds that satisfy CNF Assign a weight to each literal Weight(world) = П weights(true literals) Weighted model counting: Given CNF C and literal weights W Output Σ weights(worlds that satisfy C) PTP is reducible to lifted WMC

25. Example

27. If Then

28. Example

30. Inference Problems

31. All conditional probabilities are ratios of partition functions: All partition functions can be computed by weighted model counting Propositional Case

32. Conversion to CNF + Weights WCNF(PKB) for all (F i, Φ i ) є PKB s.t. Φ i > 0 do PKB ← PKB U {(F i  A i, 0)} \ {(F i, Φ i )} CNF ← CNF(PKB) for all ¬A i literals do W ¬Ai ← Φ i for all other literals L do w L ← 1 return (CNF, weights)

33. Probabilistic Theorem Proving PTP(PKB, Query) PKB Q ← PKB U {(Query,0)} return WMC(WCNF(PKB Q )) / WMC(WCNF(PKB))

34. Probabilistic Theorem Proving PTP(PKB, Query) PKB Q ← PKB U {(Query,0)} return WMC(WCNF(PKB Q )) / WMC(WCNF(PKB)) TP(KB, Query) KB Q ← KB U {¬ Query} return ¬SAT(CNF(KB Q )) Compare:

35. Weighted Model Counting WMC(CNF, weights) if all clauses in CNF are satisfied return if CNF has empty unsatisfied clause return 0 Base Case

36. Weighted Model Counting WMC(CNF, weights) if all clauses in CNF are satisfied return if CNF has empty unsatisfied clause return 0 if CNF can be partitioned into CNFs C 1,…, C k sharing no atoms return Decomp. Step

37. Weighted Model Counting WMC(CNF, weights) if all clauses in CNF are satisfied return if CNF has empty unsatisfied clause return 0 if CNF can be partitioned into CNFs C 1,…, C k sharing no atoms return choose an atom A return Splitting Step

38. First-Order Case PTP schema remains the same Conversion of PKB to hard CNF and weights: New atom in F i  A i is now Predicate i (variables in F i, constants in F i ) New argument in WMC: Set of substitution constraints of the form x = A, x ≠ A, x = y, x ≠ y Lift each step of WMC

39. Lifted Weighted Model Counting LWMC(CNF, substs, weights) if all clauses in CNF are satisfied return if CNF has empty unsatisfied clause return 0 Base Case

40. Lifted Weighted Model Counting LWMC(CNF, substs, weights) if all clauses in CNF are satisfied return if CNF has empty unsatisfied clause return 0 if there exists a lifted decomposition of CNF return Decomp. Step

41. Lifted Weighted Model Counting LWMC(CNF, substs, weights) if all clauses in CNF are satisfied return if CNF has empty unsatisfied clause return 0 if there exists a lifted decomposition of CNF return choose an atom A return Splitting Step

42. Extensions Unit propagation, etc. Caching / Memoization Knowledge-based model construction

43. Approximate Inference WMC(CNF, weights) if all clauses in CNF are satisfied return if CNF has empty unsatisfied clause return 0 if CNF can be partitioned into CNFs C 1,…, C k sharing no atoms return choose an atom A return with probability, etc. Splitting Step

44. MPE Inference Replace sums by maxes Use branch-and-bound for efficiency Do traceback

45. Overview Representation Inference Learning Applications

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

47. 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

48. Pseudo-Likelihood Likelihood of each variable given its neighbors in the data [Besag, 1975] 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

49. Discriminative Weight Learning Maximize conditional likelihood of query ( y ) given evidence ( x ) Expected counts can be approximated by counts in MAP state of y given x No. of true groundings of clause i in data Expected no. true groundings according to model

50. w i ← 0 for t ← 1 to T do y MAP ← Viterbi(x) w i ← w i + η [count i (y Data ) – count i (y MAP )] return ∑ t w i / T Voted Perceptron Originally proposed for training HMMs discriminatively [Collins, 2002] Assumes network is linear chain

51. HMMs are special case of MLNs Replace Viterbi by prob. theorem proving Network can now be arbitrary graph w i ← 0 for t ← 1 to T do y MAP ← PTP(MLN U {x}, y) w i ← w i + η [count i (y Data ) – count i (y MAP )] return ∑ t w i / T Voted Perceptron for MLNs

52. 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

53. 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 [Kok & Domingos, 2005] Bottom-up [Mihalkova & Mooney, 2007] Relational pathfinding [Kok & Domingos, 2009, 2010]

54. Alchemy Open-source software including: Full first-order logic syntax MAP and marginal/conditional inference Generative & discriminative weight learning Structure learning Programming language features alchemy.cs.washington.edu

55. AlchemyPrologBUGS Represent- ation F.O. Logic + Markov nets Horn clauses Bayes nets InferenceProbabilistic thm. proving Theorem proving Gibbs sampling LearningParameters & structure NoParams. UncertaintyYesNoYes RelationalYes No

56. Overview Representation Inference Learning Applications

57. Applications to Date Natural language processing Information extraction Entity resolution Link prediction Collective classification Social network analysis Robot mapping Activity recognition Scene analysis Computational biology Probabilistic Cyc Personal assistants Etc.

58. 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.

59. 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

60. 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.

61. 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.

62. 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

63. 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)

64. 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

65. 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)

66. 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

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) 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. 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”)

71. 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

72. 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

73. 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”)

74. 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”)

75. Results: Segmentation on Cora

76. Results: Matching Venues on Cora

77. Summary Cognitive modeling requires combination of logical and statistical techniques We need to unify the two Markov logic Syntax: Weighted logical formulas Semantics: Markov network templates Inference: Probabilistic theorem proving Learning: Statistical inductive logic programming Many applications to date

78. Resources Open-source software/Web site: Alchemy Learning and inference algorithms Tutorials, manuals, etc. MLNs, datasets, etc. Publications Book: Domingos & Lowd, Markov Logic, Morgan & Claypool, 2009. alchemy.cs.washington.edu