Bayesian Networks
-
Upload
graham-cleveland -
Category
Documents
-
view
21 -
download
0
description
Transcript of Bayesian Networks
![Page 1: Bayesian Networks](https://reader031.fdocuments.in/reader031/viewer/2022020423/5681348c550346895d9b73a3/html5/thumbnails/1.jpg)
Daphne Koller
Independencies
BayesianNetworks
ProbabilisticGraphicalModels
Representation
![Page 2: Bayesian Networks](https://reader031.fdocuments.in/reader031/viewer/2022020423/5681348c550346895d9b73a3/html5/thumbnails/2.jpg)
Daphne Koller
Independence & Factorization
• Factorization of a distribution P implies independencies that hold in P
• If P factorizes over G, can we read these independencies from the structure of G?
P(X,Y,Z) 1(X,Z) 2(Y,Z) P(X,Y) = P(X) P(Y) X,Y independent
(X Y | Z)
![Page 3: Bayesian Networks](https://reader031.fdocuments.in/reader031/viewer/2022020423/5681348c550346895d9b73a3/html5/thumbnails/3.jpg)
Daphne Koller
Flow of influence & d-separation
Definition: X and Y are d-separated in G given Z if there is no active trail in G between X and Y given Z
ID
G
L
S
Notation: d-sepG(X, Y | Z)
![Page 4: Bayesian Networks](https://reader031.fdocuments.in/reader031/viewer/2022020423/5681348c550346895d9b73a3/html5/thumbnails/4.jpg)
Daphne Koller
Factorization Independence: BNsTheorem: If P factorizes over G, and d-sepG(X, Y | Z) then P satisfies (X Y | Z)
ID
G
L
S
![Page 5: Bayesian Networks](https://reader031.fdocuments.in/reader031/viewer/2022020423/5681348c550346895d9b73a3/html5/thumbnails/5.jpg)
Daphne Koller
Any node is d-separated from its non-descendants given its parents IntelligenceDifficulty
Grade
Letter
SAT
JobHappy
Coherence
If P factorizes over G, then in P, anyvariable is independent of its non-descendants given its parents
![Page 6: Bayesian Networks](https://reader031.fdocuments.in/reader031/viewer/2022020423/5681348c550346895d9b73a3/html5/thumbnails/6.jpg)
Daphne Koller
I-maps• d-separation in G P satisfies
corresponding independence statement
• Definition: If P satisfies I(G), we say that G is an I-map (independency map) of P
I(G) = {(X Y | Z) : d-sepG(X, Y | Z)}
![Page 7: Bayesian Networks](https://reader031.fdocuments.in/reader031/viewer/2022020423/5681348c550346895d9b73a3/html5/thumbnails/7.jpg)
Daphne Koller
I-mapsI D Pro
bi0 d0 0.4
2i0 d1 0.1
8 i1 d0 0.2
8i1 d1 0.1
2
I D Prob.i0 d0 0.282i0 d1 0.02 i1 d0 0.564i1 d1 0.134
G1
P2
IDID
P1
G2
![Page 8: Bayesian Networks](https://reader031.fdocuments.in/reader031/viewer/2022020423/5681348c550346895d9b73a3/html5/thumbnails/8.jpg)
Daphne Koller
Factorization Independence: BNsTheorem: If P factorizes over G, then G
is an I-map for P
![Page 9: Bayesian Networks](https://reader031.fdocuments.in/reader031/viewer/2022020423/5681348c550346895d9b73a3/html5/thumbnails/9.jpg)
Daphne Koller
Independence FactorizationTheorem: If G is an I-map for P, then P
factorizes over G
ID
G
L
S
![Page 10: Bayesian Networks](https://reader031.fdocuments.in/reader031/viewer/2022020423/5681348c550346895d9b73a3/html5/thumbnails/10.jpg)
Daphne Koller
SummaryTwo equivalent views of graph structure: • Factorization: G allows P to be represented• I-map: Independencies encoded by G hold in P
If P factorizes over a graph G, we can read from the graph independencies that must hold in P (an independency map)