1. Graphical Models - Modeling - Wolfram Burgard, Luc De Raedt, Kristian Kersting, Bernhard Nebel Albert-Ludwigs University Freiburg, Germany PCWP CO HRBP HREKG HRSAT ERRCAUTER HR HISTORY CATECHOL SAO2 EXPCO2 ARTCO2 VENTALV VENTLUNG VENITUBE DISCONNECT MINVOLSET VENTMACH KINKEDTUBE INTUBATIONPULMEMBOLUS PAPSHUNT ANAPHYLAXIS MINOVL PVSAT FIO2 PRESS INSUFFANESTHTPR LVFAILURE ERRBLOWOUTPUT STROEVOLUMELVEDVOLUME HYPOVOLEMIA CVP BP Mainly based on F. V. Jensen, „Bayesian Networks and Decision Graphs“, Springer-Verlag New York, 2001. Advanced I WS 06/07

2. Bayesian Networks Advanced I WS 06/07 Outline Introduction Reminder: Probability theory Basics of Bayesian Networks Modeling Bayesian networks Inference Excourse: Markov Networks Learning Bayesian networks Relational Models

3. Bayesian Networks Advanced I WS 06/07 (Discrete) Random Variables and Probability A random variable X describes a value that is the result of a process which can not be determined in advance: the rank of the card we draw from a deck The sample space S of a random variable is the set of all possible values of the variable: Ace, King, Queen, Jack, … An event is a subset of S: {Ass, King}, {Queen}, … The probability function P(X=x), short P(x),defines the probability that X yields a certain value x. - Reminder: Probability Theory

4. Bayesian Networks Advanced I WS 06/07 (Discrete) Random Variables and Probability The world  is described as a set of random variables and divided it into elementary events or states States are mutually exclusive A probability space has the following properties: - Reminder: Probability Theory

5. Bayesian Networks Advanced I WS 06/07 Conditional probability and Bayes’ Rule P(x|y) denotes the conditional probability that x will happen given that y has happened. Bayes’ rule states that: - Reminder: Probability Theory If it is known that a certain event occurred this may change the probability that another event will occur

6. Bayesian Networks Advanced I WS 06/07 Fundamental rule The fundamental rule (sometimes called ‚chain rule‘) for probability calculus is - Reminder: Probability Theory

7. Bayesian Networks Advanced I WS 06/07 Conditional probability and Bayes’ Rule The complete probability formula states that Note: mirrors our intuition that the unconditional probability of an event is somewhere between its conditional probability based on two opposing assumptions. - Reminder: Probability Theory

8. Bayesian Networks Advanced I WS 06/07 Joint Probability Distribution „truth table“ of set of random variables Any probability we are interested in can be computed from it true1green0.001 true1blue0.021 true2green0.134 true2blue0.042... false2blue0.2 - Reminder: Probability Theory

9. Bayesian Networks Advanced I WS 06/07 Bayes’ is fundamental in learning when viewed in terms of evidence and hypothesis: Bayes´rule and Inference prior probability posterior probability likelihood This allows us to update our belief in a hypothesis based on prior belief and in response to evidence. - Reminder: Probability Theory

10. Bayesian Networks Advanced I WS 06/07 Independence among random variables Formally for a probability distribution P: - Reminder: Probability Theory Two random variables X and Y are independent if knowledge about X does not change the uncertainty about Y and vice versa.

11. Bayesian Networks Advanced I WS 06/07 Independence among random variables - Reminder: Probability Theory

12. Bayesian Networks Advanced I WS 06/07 Outline Introduction Reminder: Probability theory Basics of Bayesian Networks Modeling Bayesian networks Inference Excourse: Markov Networks Learning Bayesian networks Relational Models

13. Bayesian Networks Advanced I WS 06/07 Bayesian Networks 1.Finite, acyclic graph 2.Nodes: (discrete) random variables 3.Edges: direct influences 4.Associated with each node: a table representing a conditional probability distribution (CPD), quantifying the effect the parents have on the node MJ E B A - Bayesian Networks

14. Bayesian Networks Advanced I WS 06/07 Associated CPDs naive representation –tables other representations –decision trees –rules –neural networks –support vector machines –... E B A.9.1 e b e.7.3.99.01.8.2 be b b e BEP(A | E,B) - Bayesian Networks

15. Bayesian Networks Advanced I WS 06/07 Bayesian Networks X1X1 X2X2 X3X3 (0.2, 0.8) (0.6, 0.4) true1(0.2,0.8) true2(0.5,0.5) false1(0.23,0.77) false2(0.53,0.47) - Bayesian Networks

16. Bayesian Networks Advanced I WS 06/07 Conditional Independence I - Bayesian Networks Each node / random variable is (conditionally) independent of all its non- descendants given a joint state of its parents.

17. Bayesian Networks Advanced I WS 06/07 Example Visit to Asia? Smoker? Has bronchitisHas lung cancerHas tuberculosis Dyspnoea? Positive X-ray? Tuberculosis or cancer - Bayesian Networks

18. Bayesian Networks Advanced I WS 06/07 Example Visit to Asia? Smoker? Has bronchitis Has lung cancerHas tuberculosis Dyspnoea? Positive X-ray? Tuberculosis or cancer - Bayesian Networks

19. Bayesian Networks Advanced I WS 06/07 Example Visit to Asia? Smoker? Has bronchitisHas lung cancerHas tuberculosis Dyspnoea? Positive X-ray? Tuberculosis or cancer - Bayesian Networks

20. Bayesian Networks Advanced I WS 06/07 Example Visit to Asia? Smoker? Has bronchitis Has lung cancerHas tuberculosis Dyspnoea? Positive X-ray? Tuberculosis or cancer - Bayesian Networks

21. Bayesian Networks Advanced I WS 06/07 Summary Semantics Compact & natural representation: nodes have  k parents   2 k n instead of 2 n params conditional independencies in BN structure + local probability models full joint distribution over domain = X I S TB - Bayesian Networks

22. Bayesian Networks Advanced I WS 06/07 Operations Inference: Most probable explanation: –Most probable configuration of given the evidence Data Conflict (coherence of evidence) –positive conf(e) indicates a possible conflict Sensitivity analysis –Which evidence is in favour of/against/irrelevant for H –Which evidence discriminates H from H´ - Bayesian Networks

23. Bayesian Networks Advanced I WS 06/07 Conditional Independence II One could graphically read off other conditional independencies based on –Serial connections –Diverging connections –Converging connections ABC A B C... E A BC E - Bayesian Networks

24. Bayesian Networks Advanced I WS 06/07 Serial Connections - Example Initial opponent´s hand Opponent´s hand after the first change of cards Final opponent´s hand Poker: If we do not know the opponent´s hand after the first change of cards, knowing her initial hand will tell us more about her final hand. - Bayesian Networks

25. Bayesian Networks Advanced I WS 06/07 Serial Connections - Example Initial opponent´s hand Opponent´s hand after the first change of cards Final opponent´s hand Poker: If we do not know the opponent´s hand after the first change of cards, knowing her initial hand will tell us more about her final hand. pair - Bayesian Networks

26. Bayesian Networks Advanced I WS 06/07 Serial Connections - Example Initial opponent´s hand Opponent´s hand after the first change of cards Final opponent´s hand If we know the opponent´s hand after the first change of cards, knowing her initial hand will tell us nothing about her final hand. - Bayesian Networks

27. Bayesian Networks Advanced I WS 06/07 Serial Connections - Example Initial opponent´s hand Opponent´s hand after the first change of cards Final opponent´s hand pair flush pair If we know the opponent´s hand after the first change of cards, knowing her initial hand will tell us nothing about her final hand. - Bayesian Networks

28. Bayesian Networks Advanced I WS 06/07 Serial Connections ABC A has influence on B, and B has influence on C - Bayesian Networks

29. Bayesian Networks Advanced I WS 06/07 Serial Connections A BC A has influence on B, and B has influence on C a - Bayesian Networks

30. Bayesian Networks Advanced I WS 06/07 Serial Connections AB C A has influence on B, and B has influence on C a a - Bayesian Networks

31. Bayesian Networks Advanced I WS 06/07 Serial Connections A B C A has influence on B, and B has influence on C b - Bayesian Networks

32. Bayesian Networks Advanced I WS 06/07 Serial Connections A BC A has influence on B, and B has influence on C b b - Bayesian Networks

33. Bayesian Networks Advanced I WS 06/07 Serial Connections ABC A has influence on B, and B has influence on C Evidence on A will influence the certainty of C (through B) and vice versa - Bayesian Networks

34. Bayesian Networks Advanced I WS 06/07 Serial Connections A BC A has influence on B, and B has influence on C Evidence on A will influence the certainty of C (through B) and vice versa a - Bayesian Networks

35. Bayesian Networks Advanced I WS 06/07 Serial Connections A BC A has influence on B, and B has influence on C Evidence on A will influence the certainty of C (through B) and vice versa a a - Bayesian Networks

36. Bayesian Networks Advanced I WS 06/07 Serial Connections ABC A has influence on B, and B has influence on C Evidence on A will influence the certainty of C (through B) and vice versa aa a - Bayesian Networks

37. Bayesian Networks Advanced I WS 06/07 Serial Connections ABC IF the state of B is known, THEN the channel through B is blocked, A and B become independent A has influence on B, and B has influence on C Evidence on A will influence the certainty of C (through B) and vice versa - Bayesian Networks

38. Bayesian Networks Advanced I WS 06/07 Serial Connections A B C IF the state of B is known, THEN the channel through B is blocked, A and B become independent A has influence on B, and B has influence on C Evidence on A will influence the certainty of C (through B) and vice versa b - Bayesian Networks

39. Bayesian Networks Advanced I WS 06/07 Serial Connections A B C IF the state of B is known, THEN the channel through B is blocked, A and B become independent A has influence on B, and B has influence on C Evidence on A will influence the certainty of C (through B) and vice versa b a a - Bayesian Networks

40. Bayesian Networks Advanced I WS 06/07 Conditional Independence II One could graphically read off other conditional independencies based on –Serial connections –Diverging connections –Converging connections ABC A B C... E A BC E - Bayesian Networks

41. Bayesian Networks Advanced I WS 06/07 Diverging Connections - Example Stature Hair length Sex If we do not know the sex of a person, seeing the length of her hair will tell us more about the sex, and this in turn will focus our belief on her stature. - Bayesian Networks

42. Bayesian Networks Advanced I WS 06/07 Diverging Connections - Example Stature Hair length Sex long If we do not know the sex of a person, seeing the length of her hair will tell us more about the sex, and this in turn will focus our belief on her stature. - Bayesian Networks

43. Bayesian Networks Advanced I WS 06/07 Diverging Connections - Example Stature Hair length Sex If we know that the person is a man, then the length of hair gives us no extra clue on his stature. - Bayesian Networks

44. Bayesian Networks Advanced I WS 06/07 Diverging Connections - Example Stature Hair length Sex long male If we know that the person is a man, then the length of hair gives us no extra clue on his stature. - Bayesian Networks

45. Bayesian Networks Advanced I WS 06/07 Diverging Connections A B C... E Influence can pass between all the children of A unless the state of A is known - Bayesian Networks

46. Bayesian Networks Advanced I WS 06/07 Diverging Connections A B C... E Influence can pass between all the children of A unless the state of A is known b - Bayesian Networks

47. Bayesian Networks Advanced I WS 06/07 Diverging Connections A B C... E Influence can pass between all the children of A unless the state of A is known b b - Bayesian Networks

48. Bayesian Networks Advanced I WS 06/07 Diverging Connections A B C... E Influence can pass between all the children of A unless the state of A is known b b b - Bayesian Networks

49. Bayesian Networks Advanced I WS 06/07 Diverging Connections A B C... E Influence can pass between all the children of A unless the state of A is known bb b - Bayesian Networks

50. Bayesian Networks Advanced I WS 06/07 Diverging Connections A B C... E Influence can pass between all the children of A unless the state of A is known a - Bayesian Networks

51. Bayesian Networks Advanced I WS 06/07 Diverging Connections A B C... E Influence can pass between all the children of A unless the state of A is known a b - Bayesian Networks

52. Bayesian Networks Advanced I WS 06/07 Diverging Connections A B C... E Influence can pass between all the children of A unless the state of A is known a b b - Bayesian Networks

53. Bayesian Networks Advanced I WS 06/07 Conditional Independence II One could graphically read off other conditional independencies based on –Serial connections –Diverging connections –Converging connections ABC A B C... E A BC E - Bayesian Networks

54. Bayesian Networks Advanced I WS 06/07 Converging Connections - Example Flue Salmonella Pallor Nausea If we know nothing of nausea or pallor, then the information on whether the person has a Salmonella infection will tell us anything about flu. - Bayesian Networks

55. Bayesian Networks Advanced I WS 06/07 Converging Connections - Example Flue Salmonella Pallor Nausea yes If we know nothing of nausea or pallor, then the information on whether the person has a Salmonella infection will tell us anything about flu. - Bayesian Networks

56. Bayesian Networks Advanced I WS 06/07 Converging Connections - Example Flu Salmonella Pallor Nausea If we have noticed the person is pale, then the information that the she does not have a Salmonella infection will make us more ready to believe that she has the flu. - Bayesian Networks

57. Bayesian Networks Advanced I WS 06/07 Converging Connections - Example Flu Salmonella Pallor Nausea yes If we have noticed the person is pale, then the information that the she does not have a Salmonella infection will make us more ready to believe that she has the flu. - Bayesian Networks

58. Bayesian Networks Advanced I WS 06/07 Converging Connections A B C... E Evidence of one of A´s parents has no influence on the certainty of the others - Bayesian Networks

59. Bayesian Networks Advanced I WS 06/07 Converging Connections A B C... E Evidence of one of A´s parents has no influence on the certainty of the others b - Bayesian Networks

60. Bayesian Networks Advanced I WS 06/07 Converging Connections A B C... E Evidence of one of A´s parents has no influence on the certainty of the others b b - Bayesian Networks

61. Bayesian Networks Advanced I WS 06/07 Converging Connections A B C... E Evidence of one of A´s parents has no influence on the certainty of the others b b b - Bayesian Networks

62. Bayesian Networks Advanced I WS 06/07 Converging Connections A B C... E Evidence of one of A´s parents has no influence on the certainty of the others b b b b - Bayesian Networks

63. Bayesian Networks Advanced I WS 06/07 Converging Connections A B C... E However, if anything is known about the consequence, then information on one possible cause may tell us something about the other causes. (explaining away) - Bayesian Networks

64. Bayesian Networks Advanced I WS 06/07 Converging Connections A B C... E However, if anything is known about the consequence, then information on one possible cause may tell us something about the other causes. (explaining away) a - Bayesian Networks

65. Bayesian Networks Advanced I WS 06/07 Converging Connections A B C... E However, if anything is known about the consequence, then information on one possible cause may tell us something about the other causes. (explaining away) a b - Bayesian Networks

66. Bayesian Networks Advanced I WS 06/07 Converging Connections A B C... E However, if anything is known about the consequence, then information on one possible cause may tell us something about the other causes. (explaining away) a b b - Bayesian Networks

67. Bayesian Networks Advanced I WS 06/07 Converging Connections A B C... E However, if anything is known about the consequence, then information on one possible cause may tell us something about the other causes. (explaining away) a bb b - Bayesian Networks

68. Bayesian Networks Advanced I WS 06/07 Converging Connections A B C... E However, if anything is known about the consequence, then information on one possible cause may tell us something about the other causes. (explaining away) a b b b - Bayesian Networks

69. Bayesian Networks Advanced I WS 06/07 Two distinct variables A and B in a Bayesian network are d-separated if, for all (undirected) paths between A and B, there is an intermediate variable V (distinct from A and B) such that The connection is serial or diverging and V is instantiated, or The connection is converging, and neither V nor any of V´s descendants have reveived evidence. If A and B are not d-separated, then they are called d-connected. Conitional Independence II: d-Separation - Bayesian Networks

70. Bayesian Networks Advanced I WS 06/07 d-Separation If A and B are d-separated through the intermediate variable V then (if the connection is serial or diverging and V is instantiated) (if the connection is converging, and Neither V nor any of V´s descendants have received evidence) - Bayesian Networks

71. Bayesian Networks Advanced I WS 06/07 Outline Introduction Reminder: Probability theory Bascis of Bayesian Networks Modeling Bayesian networks Inference Markov Networks Learning Bayesian networks Relational Models

72. Bayesian Networks Advanced I WS 06/07 Undirected Relations Dependence relation R among A,B,C Neither desirable nor possible to attach directions to them A BC - Modeling D A B C Add new variable D with states y, n Set Enter evidence D=y

73. Bayesian Networks Advanced I WS 06/07 Noisy or When a variable A has several parents, you must specify the CPD for each configuration of its parents –Number of cases for each configuration in a database to small, configurations to specific for experts,... Sore throat Cold Agina ? - Modeling

74. Bayesian Networks Advanced I WS 06/07 Noisy or binary variables listing all the causes of binary variable Each event causes (independently of the other events) unless an inhibitor prevents it with probability Then where Y is the set of indeces of variables in state y - Modeling

75. Bayesian Networks Advanced I WS 06/07 Noisy or linear in the number of parents Can be modelled directly: Complementary construction is noisy and A1A1 A2A2 AkAk... B1B1 B2B2 BkBk B logical or - Modeling

76. Bayesian Networks Advanced I WS 06/07 Divorcing A1A1 A2A2 AnAn... B - Modeling A3A3

77. Bayesian Networks Advanced I WS 06/07 Divorcing A1A1 A2A2 AnAn... B - Modeling A3A3 A1A1 A2A2 AnAn... B C s 1,...,s m A3A3

78. Bayesian Networks Advanced I WS 06/07 Divorcing A1A1 A2A2 AnAn... B - Modeling The set of parents A 1,...,A i for B are divorced from the parents A i+1,...,A n Set of configurations c 1,...,c k of A 1,...,A i can be partitioned into the sets s 1,...,s m s.t. whenever two configurations c,c´ are elements of the same s i then A3A3 A1A1 A2A2 AnAn... B C s 1,...,s m A3A3

79. Bayesian Networks Advanced I WS 06/07 Expert e 1,...,e n disagree on the conditional probabilities of a Bayesian network Experts disagree on A, D Confidences into experts c 1,...,c n e.g. (2, 1, 5, 3,..., 1) A C D B - Modeling Experts Disagreements

80. Bayesian Networks Advanced I WS 06/07 Introduce a new variable E having the (confidence) distribution p 1,...,p n and link it to each variable, the tables of which the experts disagree upon. The child variables have a CPD for each expert Expert disagreements A C D B E e 1,...,e N - Modeling

81. Bayesian Networks Advanced I WS 06/07 Other Models … - Modeling