Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with...
Transcript of Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with...
![Page 1: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/1.jpg)
Problem Solving with Graphical Models
Rina DechterDonald Bren School of Computer Science
University of California, Irvine, USA
Dagshtul, 2011
![Page 2: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/2.jpg)
Radcliffe2
What is Artificial Intelligence(John McCarthy , Basic Questions)
What is artificial intelligence? It is the science and engineering of making intelligent machines,
especially intelligent computer programs. It is related to the similar task of using computers to understand human intelligence, but AI does not have to confine itself to methods that are biologically observable.
Yes, but what is intelligence? Intelligence is the computational part of the ability to achieve goals in
the world. Varying kinds and degrees of intelligence occur in people, many animals and some machines.
Isn't there a solid definition of intelligence that doesn't depend on relating it to human intelligence?
Not yet. The problem is that we cannot yet characterize in general what kinds of computational procedures we want to call intelligent. We understand some of the mechanisms of intelligence and not others.
More in: http://www-formal.stanford.edu/jmc/whatisai/node1.html
![Page 3: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/3.jpg)
3
Mechanical Heuristic generation
Observation: People generate heuristics by consulting simplified/relaxed models.Context: Heuristic search (A*) of state-space graph (Nillson, 1980)Context: Weak methods vs. strong methodsDomain knowledge: Heuristic function
h(n):Heuristic underestimatethe best cost fromn to the solution
![Page 4: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/4.jpg)
4
The Simplified models Paradigm
Pearl 1983 (On the discovery and generation of certain Heuristics, 1983, AI Magazine, 22-23) : “knowledge about easy problems could serve as a heuristic in the solution of difficult problems, i.e., that it should be possible to manipulate the representation of a difficult problem until it is approximated by an easy one, solve the easy problem, and then use the solution to guide the search process in the original problem.”
The implementation of this scheme requires three major steps:a) simplification, b) solution, andc) advice generation.
Simplified = relaxed is appealing because:1. implies admissibility, monotonicity,2. explains many human-generated heuristics (15-puzzle, traveling salesperson)
“We must have a simple a-priori criterion for deciding when a problem lends itself to easy solution.”
![Page 5: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/5.jpg)
5
Systematic relaxation of STRIPS
STRIPS (Stanford Research Institute Problem Solver, Nillson and Fikes 1971) action representation:
Move(x,c1,c2)Precond list: on(x1,c1), clear(c2), adj(c1,c2)
Add-list: on(x1,c2), clear(c1)Delete-list: on(x1,c1), clear(c2)
Relaxation (Sacerdoti, 1974): Remove literals from the precondition-list: 1. clear(c2), adj(c2,c3) #misplaced tiles2. Remove clear(c2) manhatten distance3. Remove adj(c2,c3) h3, a new procedure that transfer to the empty location a tile appearing there in the goal
But the main question remained:“Can a program tell an easy problem from a hard one without actually solving?” (Pearl 1984, Heuristics)
![Page 6: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/6.jpg)
6
Easy = Greedily solved?
Pearl, 84: Most easy problems we encounter are solved by “greedy” hill-climbing methods without backtracking” and that the features that make them amenable to such methods is their “decomposability”
The question now:Can we recognize a greedily solved STRIPS problem?”
![Page 7: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/7.jpg)
7
Whow! Backtrack-free is greedy!I read Montanari (1974),I read mackworth, (1977)Got absorbed…
Freuder, JACM 1982 : “A sufficient condition for backtrack-free search”
Sufficient condition (Freuder 82):1. Trees (width-1) and arc-consistency implies backtrack-free2. Width=i and (i+1)-consistency implies backtrack-free search
Arc-consistentNo dead-ends
This moved me to constraint network and ultimately to graphical models.But: Is it indeed the case that heuristics are generated by simplifiedModels?
If 3-consistent no deadends
W=1W=2
![Page 8: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/8.jpg)
Outline of the talk
Introduction to graphical models Inference: Exact and approximate Conditioning Search: exact and approximate Hybrids of search and inference (exact) Compilation, (e.g., AND/OR Decision Diagrams) Questions:
Representation issues: directed vs undirected The role of hidden variables Finding good structure How can we predict problem instance hardness?
![Page 9: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/9.jpg)
Outline
What are graphical models Overview of Inference Search and their hybrids
Inference: Exact and approximate Conditioning Search: exact and approximate Hybrids of search and inference (exact) Compilation, (e.g., AND/OR Decision Diagrams) Questions:
Representation issues: directed vs undirected The role of hidden variables Finding good structure Representation guided by human representation Computation: inspired by human thinking
![Page 10: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/10.jpg)
What are Graphical Models
A way to represent global knowledge, mostly declaratively, using small local pieces of functions/relations. Combined, they give a global view of a world about which we want to reason, namely to answer queries.
Different types of graphs can capture variable interaction through the local functions.
Because representation is modular, reasoning can be modular too.
![Page 11: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/11.jpg)
11
A Bred greenred yellowgreen redgreen yellowyellow greenyellow red
Map coloring
Variables: countries (A B C etc.)
Values: colors (red green blue)
Constraints: ... , ED D, AB,A
C
A
B
DE
F
G
Constraint Networks
Constraint graph
A
BD
CG
F
E
Global view: all solutions.Tasks: Is there a solution?, find one, fine all, count all
![Page 12: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/12.jpg)
Three dimensional interpretation of 2 dimentional drawing
![Page 13: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/13.jpg)
Huffman-Clowes junction labelings (1975)
![Page 14: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/14.jpg)
Sudoku –Constraint Satisfaction
Each row, column and major block must be alldifferent
“Well posed” if it has unique solution: 27 constraints
2 34 62
•Variables: empty slots
•Domains = {1,2,3,4,5,6,7,8,9}
•Constraints: •27 all-different
•Constraint •Propagation
•Inference
![Page 15: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/15.jpg)
Bayesian Networks(Pearl, 1988)
Gloabl view: P(S, C, B, X, D) = P(S) P(C|S) P(B|S) P(X|C,S) P(D|C,B)
lung Cancer
Smoking
X-ray
Bronchitis
DyspnoeaP(D|C,B)
P(B|S)
P(S)
P(X|C,S)
P(C|S)
Θ) (G,BN
CPD:C B P(D|C,B)0 0 0.1 0.90 1 0.7 0.31 0 0.8 0.21 1 0.9 0.1
Belief Updating, Most probable tuple (MPE)
= find argmax P(S)· P(C|S)· P(B|S)· P(X|C,S)· P(D|C,B) =?
![Page 16: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/16.jpg)
The “alarm” network - 37 variables, 509 parameters (instead of 237)
PCWP CO
HRBP
HREKG HRSAT
ERRCAUTERHRHISTORY
CATECHOL
SAO2 EXPCO2
ARTCO2
VENTALV
VENTLUNG VENITUBE
DISCONNECT
MINVOLSET
VENTMACHKINKEDTUBEINTUBATIONPULMEMBOLUS
PAP SHUNT
ANAPHYLAXIS
MINOVL
PVSAT
FIO2PRESS
INSUFFANESTHTPR
LVFAILURE
ERRBLOWOUTPUTSTROEVOLUMELVEDVOLUME
HYPOVOLEMIA
CVP
BP
Monitoring Intensive-Care Patients
![Page 17: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/17.jpg)
Mixed Probabilistic and Deterministic networks
P(C|W)P(B|W)
P(W)
P(A|W)
W
B A C
Query:Is it likely that Chris goes to the party if Becky does not but the weather is bad?
PN CN
Semantics?
Algorithms?
),,|,( ACBAbadwBCP
A→B C→AB A CP(C|W)P(B|W)
P(W)
P(A|W)
W
B A C
A→B C→AB A C
Alex is-likely-to-go in bad weatherChris rarely-goes in bad weatherBecky is indifferent but unpredictable
If Alex goes, then Becky goes:If Chris goes, then Alex goes:
W A P(A|W)
good 0 .01
good 1 .99
bad 0 .1
bad 1 .9
![Page 18: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/18.jpg)
Applications
Markovnetworks
![Page 19: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/19.jpg)
A graphical model (X,D,F): X = {X1,…Xn} variables D = {D1, … Dn} domains F = {f1,…,fm} functions
Operators: combination elimination (projection)
Graphical Models
)( : CAFfi
A
D
BC
E
F
A C F P(F|A,C)0 0 0 0.140 0 1 0.960 1 0 0.400 1 1 0.601 0 0 0.351 0 1 0.651 1 0 0.721 1 1 0.68
Primal graph(interaction graph)
A C Fred green blueblue red redblue blue green
green red blue
Relation
A
D
BC
E
F
![Page 20: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/20.jpg)
Complexity of Reasoning Tasks Constraint satisfaction Counting solutions Combinatorial optimization Belief updating Most probable explanation Decision-theoretic planning
Reasoning iscomputationally hard
Linear / Polynomial / Exponential
0
200
400
600
800
1000
1200
1 2 3 4 5 6 7 8 9 10
n
f(n)LinearPolynomialExponential
Complexity isTime and space(memory) exponential
A
D
BC
E
F
![Page 21: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/21.jpg)
Tree-solving is easy
Belief updating (sum-prod)
MPE (max-prod)
CSP – consistency (projection-join)
#CSP (sum-prod)
P(X)
P(Y|X) P(Z|X)
P(T|Y) P(R|Y) P(L|Z) P(M|Z)
)(XmZX
)(XmXZ
)(ZmZM)(ZmZL
)(ZmMZ)(ZmLZ
)(XmYX
)(XmXY
)(YmTY
)(YmYT
)(YmRY
)(YmYR
Trees are processed in linear time and memory
![Page 22: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/22.jpg)
Counting
1 2 3 4
4 3 2 155
5 5 5
How many people?
SUM operatorCHAIN structure
![Page 23: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/23.jpg)
Maximization
What is the maximum?
15
23
10 32
10
100
65
47
50
77
100
15
23
77
10047
100
77
100
100
100
10023
77
10
100
100
32
MAX operatorTREE structure
![Page 24: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/24.jpg)
12” 14” 15”
S
I II III
P60G 80G
H
6C 9C
B
Min-Cost Assignment
What is minimum cost configuration?
6C 9C
I 30 50
II 40 55
III ∞ 60
I II III
12” 45 ∞ ∞
14” 50 60 70
15” ∞ 65 8060G 80G
12” 30 50
14” 40 45
15” 50 ∞
I II III
12” 75 ∞ ∞
14” 80 100 130
15” ∞ 105 140
12” 14” 15”
105 120 155
40
II
30
I
60
III+
MIN-SUM operatorsCHAIN structure
60
40
30
105
80
75
80
14”
75
12”
105
15”
50
40
30
40
14”
30
12”
50
15”
+ =
![Page 25: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/25.jpg)
Belief Updating
Buzzsound
Mechanical problem
Hightemperature
Faultyhead
Readdelays
H P(H)0 .91 .1
F P(F)0 .991 .01
H F M P(M|H,F)0 0 0 .90 0 1 .10 1 0 .10 1 1 .91 0 0 .81 0 1 .21 1 0 .011 1 1 .99
F R P(R|F)0 0 .80 1 .21 0 .30 1 .7
P(F | B=1) = ?
M h1(M)0 .051 .8
H F M Bel(M,H,F)0 0 0 .04050 0 1 .0720 1 0 .00450 1 1 .6481 0 0 .0041 0 1 .0081 1 0 .000051 1 1 .0792
H h2(H)0 .91 .1
F h3(F)0 .12451 .73175
F h4(F)0 11 1
H F M P(M|H,F)0 0 0 .90 0 1 .10 1 0 .10 1 1 .91 0 0 .81 0 1 .21 1 0 .011 1 1 .99
* * =
M B P(B|M)0 0 .950 1 .051 0 .21 1 .8
* * =F P(F,B=1)0 .1232551 .073175
P(B=1) = .19643
Probability of evidence
P(F=1|B=1) = .3725
Updated belief
SUM-PROD operatorsPOLY-TREE structure
P(h,f,r,m,b) = P(h) P(f) P(m|h,f) P(r|f) P(b|m)
![Page 26: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/26.jpg)
Belief Propagation
Instances of tree message passing algorithm
Exact for trees
Linear in the input size
Importance: One of the first algorithms for inference in Bayesian networks Gives a cognitive dimension to its computations Basis for conditioning algorithms for arbitrary Bayesian network Basis for Loopy Belief Propagation (approximate algorithms)
(Pearl, 1988)
![Page 27: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/27.jpg)
Transforming into a Tree
By Inference (thinking) Transform into a single, equivalent tree of sub-
problems
By Conditioning (guessing) Transform into many tree-like sub-problems.
![Page 28: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/28.jpg)
Inference and Treewidth
EK
F
L
H
C
BA
M
G
J
D
ABC
BDEF
DGF
EFH
FHK
HJ KLM
treewidth = 4 - 1 = 3treewidth = (maximum cluster size) - 1
Inference algorithm:Time: exp(tree-width)Space: exp(tree-width)
Key parameter: w*
![Page 29: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/29.jpg)
Conditioning and Cycle cutset
C P
J A
L
B
E
DF M
O
H
K
G N
C P
J
L
B
E
DF M
O
H
K
G N
A
C P
J
L
E
DF M
O
H
K
G N
B
P
J
L
E
DF M
O
H
K
G N
C
Cycle cutset = {A,B,C}
C P
J A
L
B
E
DF M
O
H
K
G N
C P
J
L
B
E
DF M
O
H
K
G N
C P
J
L
E
DF M
O
H
K
G N
C P
J A
L
B
E
DF M
O
H
K
G N
![Page 30: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/30.jpg)
Search over the Cutset
A=yellow A=green
• Inference may require too much memory
• Condition (guessing) on some of the variables
C
B K
G
LD
FH
M
J
E
C
B K
G
LD
FH
M
J
E
AC
B K
G
LD
FH
M
J
E
GraphColoringproblem
![Page 31: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/31.jpg)
Search over the Cutset (cont)
A=yellow A=green
B=red B=blue B=red B=blueB=green B=yellow
C
K
G
LD
FH
M
J
E
C
K
G
LD
FH
M
J
E
C
K
G
LD
FH
M
J
E
C
K
G
LD
FH
M
J
E
C
K
G
LD
FH
M
J
E
C
K
G
LD
FH
M
J
E
• Inference may require too much memory
• Condition on some of the variablesA
C
B K
G
LD
FH
M
J
E
GraphColoringproblem
Key parameters:the cycle-cutset, w-cutsetComplexity: exp(cutset-size)
![Page 32: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/32.jpg)
Inference vs Conditioning-Search
Inference
exp(treewidth) time/space
A
D
B C
E
F0 1 0 1 0 1 0 1
0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1
0101010101 0101010101010101010101 0101010101010101010101 0101 010101
0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1
0 1 0 1
E
C
F
D
B
A 0 1
Searchexp(n) timeOr exp(pseudo-tree O(n) space
E K
F
L
H
C
BA
M
G
J
D
ABC
BDEF
DGF
EFH
FHK
HJ KLM
A=yellow A=green
B=blue B=red B=blueB=green
CK
G
LD
F H
M
J
EACB K
G
LD
F H
M
J
E
CK
G
LD
F H
M
J
EC
K
G
LD
F H
M
J
EC
K
G
LD
F H
M
J
ESearch+inference:Space: exp(w of sub problem)Time: exp(w+cutset(w))
![Page 33: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/33.jpg)
Approximation Algorithms Since inference, search and hybrids are too
expensive when graph is dense; (high treewidth) then:
Bounding inference: Bounding the clusters by i-bound mini-bucket(i) and bounded-i-consistency Belief propagation and constraint propagation
Bounding search: Sampling Stochastic local search
Hybrid of sampling and bounded inference
Goal: an anytime scheme
![Page 34: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/34.jpg)
Search vs. Inference
A
G
B
C
E
D
F
Search (conditioning) Inference (elimination)
A=1 A=k…
G
B
C
E
D
F
G
B
C
E
D
F
A
G
B
C
E
D
F
G
B
C
E
D
F
k “sparser” problems 1 “denser” problem
![Page 35: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/35.jpg)
Outline
What are graphical models Inference: Exact and approximate Conditioning Search: exact and approximate Hybrids of search and inference (exact) Compilation, (e.g., AND/OR Decision Diagrams) Questions:
Representation issues: directed vs undirected The role of hidden variables Finding good structure Representation guided by human representation Computation: inspired by human thinking
Complete
Incomplete
Simulated Annealing
Gradient Descent
Complete
Incomplete
Adaptive Consistency
Tree Clustering
Variable Elimination
Resolution
Local Consistency
Unit Resolution
Mini‐bucket(i)
Stochastic Local SearchDFS search
Branch‐and‐Bound
A*Hybrids
![Page 36: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/36.jpg)
Bucket E: E D, E CBucket D: D ABucket C: C BBucket B: B ABucket A:
A C
contradiction
=
D = C
B = A
Bucket Elimination, Variable elimination
=
![Page 37: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/37.jpg)
free)-(backtrack problem solved greedily a getwe d ordering along widthinduced -(d)
, *
*
w(d)))exp(w O(n :Complexity
E
D
A
C
B
}2,1{
}2,1{}2,1{
}2,1{ }3,2,1{
:)(AB :)(BC :)(AD :)(
BE C,E D,E :)(
ABucketBBucketCBucketDBucketEBucket
A
E
D
C
B
:)(EB :)(
EC , BC :)(ED :)(
BA D,A :)(
EBucketBBucketCBucketDBucketABucket
E
A
D
C
B
|| RDBE ,
|| RE
|| RDB
|| RDCB|| RACB|| RAB
RA
RCBE
Bucket EliminationAdaptive Consistency (Dechter & Pearl, 1987)
![Page 38: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/38.jpg)
Directional Resolution Bucket-elimination
))exp(( :space and timeDR))(exp(||
*
*
wnOwObucketi
(Original Davis-Putnam algorithm 1960)
![Page 39: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/39.jpg)
Belief Updating/Probability of evidencePartition function
“Moral” graph
A
D E
CB
P(a|e=0) P(a,e=0)=
bcde ,,,0
P(a)P(b|a)P(c|a)P(d|b,a)P(e|b,c)=
0e
P(a) d
),,,( ecdahB
bP(b|a)P(d|b,a)P(e|b,c)
B C
ED
Variable Elimination
P(c|a)c
![Page 40: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/40.jpg)
Bucket Elimination Algorithm elim-bel (Dechter 1996)
b
Elimination operator
P(a|e=0)
W*=4Exp(w*)
bucket B:
P(a)
P(c|a)
P(b|a) P(d|b,a) P(e|b,c)
bucket C:
bucket D:
bucket E:
bucket A:
e=0
B
C
D
E
A
e)(a,hD
(a)hE
e)c,d,(a,hB
e)d,(a,hC
![Page 41: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/41.jpg)
Finding
bmax Elimination operator
MPE
W*=4”induced width” (max clique size)
bucket B:
P(a)
P(c|a)
P(b|a) P(d|b,a) P(e|b,c)
bucket C:
bucket D:
bucket E:
bucket A:
e=0
B
C
D
E
A
e)(a,hD
(a)hE
e)c,d,(a,h B
e)d,(a,hC
)xP(maxMPEx
),|(),|()|()|()(maxby replaced is
,,,,cbePbadPabPacPaPMPE
:
bcdea max
![Page 42: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/42.jpg)
Generating the MPE-tuple
C:
E:
P(b|a) P(d|b,a) P(e|b,c)B:
D:
A: P(a)
P(c|a)
e=0 e)(a,hD
(a)hE
e)c,d,(a,hB
e)d,(a,hC
(a)hP(a)max arga' 1. E
a
0e' 2.
)e'd,,(a'hmax argd' 3. C
d
)e'c,,d',(a'h)a'|P(cmax argc' 4.
Bc
)c'b,|P(e')a'b,|P(d')a'|P(bmax argb' 5.
b
)e',d',c',b',(a' Return
![Page 43: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/43.jpg)
),|()|()(),()2,1( bacpabpapcbha
),(),|()|(),( )2,3(,
)1,2( fbhdcfpbdpcbhfd
),(),|()|(),( )2,1(,
)3,2( cbhdcfpbdpfbhdc
),(),|(),( )3,4()2,3( fehfbepfbh
e
),(),|(),( )3,2()4,3( fbhfbepfehb
),|(),()3,4( fegGpfeh e
G
E
F
C D
B
A
Time: O ( exp(w+1 ))Space: O ( exp(sep))
Cluster Tree PropagationJoin-tree clustering (Spigelhalter et. Al. 1988, Dechter, Pearl 1987)
For each cluster P(X|e) is computed
A B Cp(a), p(b|a), p(c|a,b)
B C D Fp(d|b), p(f|c,d)
B E Fp(e|b,f)
E F Gp(g|e,f)
EF
BF
BC
![Page 44: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/44.jpg)
Complexity of Elimination
))((exp ( * dwnOddw ordering along graph moral of widthinduced the)(*
The effect of the ordering:
4)( 1* dw 2)( 2
* dw“Moral” graph
A
D E
CB
B
C
D
E
A
E
D
C
B
A
Trees are easyE K
L
H
C
A
M
J
ABC
BDEF
BDFG
EFH
FHK
HJ KLM
D
G
BF
Spigelhalter et. Al. 1983, Junction tree algorithm (join-tree algorithm)
![Page 45: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/45.jpg)
Outline
What are graphical models Inference: Exact and approximate Conditioning Search: exact and approximate Hybrids of search and inference (exact) Compilation, (e.g., AND/OR Decision Diagrams) Questions:
Representation issues: directed vs undirected The role of hidden variables Finding good structure Representation guided by human representation Computation: inspired by human thinking
![Page 46: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/46.jpg)
Approximate Inference
Mini-buckets, mini-clusters, i-consistency Belief propagation, constraint propagation,
Generalized belief propagation
Complete
Incomplete
Simulated Annealing
Gradient Descent
Complete
Incomplete
Adaptive Consistency
Tree Clustering
Variable Elimination
Resolution
Local Consistency
Unit Resolution
Mini‐bucket(i)
Stochastic Local SearchDFS search
Branch‐and‐Bound
A*Hybrids
![Page 47: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/47.jpg)
From Global to Local Consistency
Leads to one pass directional bounded inference, or Iterative propagation algorithms
![Page 48: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/48.jpg)
Mini-Bucket Elimination
A
B C
D
E
P(A)
P(B|A) P(C|A)
P(E|B,C)
P(D|A,B)
Bucket B
Bucket C
Bucket D
Bucket E
Bucket A
P(B|A) P(D|A,B)P(E|B,C)
P(C|A)
E = 0
P(A)
maxB∏
hB (A,D)
MPE* is an upper bound on MPE --UGenerating a solution yields a lower bound--L
maxB∏
hD (A)
hC (A,E)
hB (C,E)
hE (A)
![Page 49: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/49.jpg)
MBE(i) (Dechter and Rish 1997)
Input: i – max number of variables allowed in a mini-bucket Output: [lower bound (P of a sub-optimal solution), upper bound]
Example: approx-mpe(3) versus elim-mpe
2* w 4* w
![Page 50: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/50.jpg)
Properties of MBE(i)/mc(I)
Complexity: O(r exp(i)) time and O(exp(i)) space. Yields an upper-bound and a lower-bound.
Accuracy: determined by upper/lower (U/L) bound.
As i increases, both accuracy and complexity increase.
Possible use of mini-bucket approximations: As anytime algorithms As heuristics in search
Other tasks: similar mini-bucket approximations for: belief updating, MAP and MEU (Dechter and Rish, 1997)
![Page 51: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/51.jpg)
Approximate Inference
Mini-buckets, mini-clusters Belief propagation, constraint propagation,
Generalized belief propagation
Complete
Incomplete
Simulated Annealing
Gradient Descent
Complete
Incomplete
Adaptive Consistency
Tree Clustering
Variable Elimination
Resolution
Local Consistency
Unit Resolution
Mini‐bucket(i)
Stochastic Local SearchDFS search
Branch‐and‐Bound
A*Hybrids
![Page 52: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/52.jpg)
32,1,
32,1, 32,1,
1 X, Y, Z, T 3X YY = ZT ZX T
X Y
T Z
32,1,
=
Iterative, directional algorithms, vs Propagation algorithmsI
Arcs-consistency
![Page 53: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/53.jpg)
1 X, Y, Z, T 3X YY = ZT ZX T
X Y
T Z
=
1 3
2 3
X YXYX DRR
Arc-consistency
![Page 54: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/54.jpg)
Iterative (Loopy) Belief Proapagation
Belief propagation is exact for poly-trees IBP - applying BP iteratively to cyclic networks
No guarantees for convergence Works well for many coding networks
)( 11uX
1U 2U 3U
2X1X
)( 12xU
)( 12uX
)( 13xU
) BEL(U update :step One
1
![Page 55: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/55.jpg)
A
ABDE
FGI
ABC
BCE
GHIJ
CDEF
FGH
C
H
A
AB BC
BE
CDE CE
H
FFG GH
GI
Collapsing Clusters
ABCDE
FGI
BCE
GHIJ
CDEF
FGH
BC
CDE CE
FFG GH
GI
ABCDE
FGI
BCE
GHIJ
CDEF
FGH
BC
CDE CE
FFG GH
GI
ABCDE
FGI
BCE
GHIJ
CDEF
FGH
BC
CDE CE
FFG GH
GI
ABCDE
FGHI GHIJ
CDEF
CDE
F
GHI
![Page 56: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/56.jpg)
Join-Graphs
A
ABDE
FGI
ABC
BCE
GHIJ
CDEF
FGH
C
H
A C
A AB BC
BE
C
CDE CE
F H
FFG GH H
GI
A
ABDE
FGI
ABC
BCE
GHIJ
CDEF
FGH
C
H
A
AB BC
CDE CE
H
FF GH
GI
ABCDE
FGI
BCE
GHIJ
CDEF
FGH
BC
DE CE
FF GH
GI
ABCDE
FGHI GHIJ
CDEF
CDE
F
GHI
more accuracy
less complexity
![Page 57: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/57.jpg)
Outline
What are graphical models Inference: Exact and approximate Conditioning Search: exact and approximate Hybrids of search and inference (exact) Compilation, (e.g., AND/OR Decision Diagrams) Questions:
Representation issues: directed vs undirected The role of hidden variables Finding good structure Representation guided by human representation Computation: inspired by human thinking
![Page 58: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/58.jpg)
Backtracking Search for a Solution
![Page 59: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/59.jpg)
Belief Updating: Searching the Probability Tree
0
),|(),|()|()|()()0,(ebcb
cbePbadPacPabPaPeaP
Brute-force Complexity: O(exp(n)), linear spaceSame as counting solutions
![Page 60: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/60.jpg)
OR search space
A
D
B C
E
F
0 1 0 1 0 1 0 1
0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1
0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1
0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1
0 1 0 1
E
C
F
D
B
A 0 1
Ordering: A B E C D F
Constraint network
Size of search space: O(exp n)
![Page 61: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/61.jpg)
AND/OR Search Space
AOR
0AND 1
BOR B
0AND 1 0 1
EOR C E C E C E C
OR D F D F D F D F D F D F D F D F
AND 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1
AND 0 10 1 0 10 1 0 10 1 0 10 1
A
D
B C
E
F
A
D
B
CE
F
Primal graph DFS tree
A
D
B C
E
F
A
D
B C
E
F
![Page 62: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/62.jpg)
AND/OR vs. ORAOR
0AND 1
BOR B
0AND 1 0 1
EOR C E C E C E C
OR D F D F D F D F D F D F D F D F
AND 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1
AND 0 10 1 0 10 1 0 10 1 0 10 1
E 0 1 0 1 0 1 0 1
0C 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1
F 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1
D 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1
0B 1 0 1
A 0 1
E 0 1 0 1 0 1 0 1
0C 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1
F 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1
D 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1
0B 1 0 1
A 0 1
AND/OR
OR
A
D
B C
E
FA
D
B
CE
F
1
1
1
0
1
0
AND/OR size: exp(4), OR size exp(6)
![Page 63: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/63.jpg)
Pseudo-Trees(Freuder 85, Bayardo 95, Bodlaender and Gilbert, 91)
(a) Graph
4 61
3 2 7 5
(b) DFS treedepth=3
(c) pseudo- treedepth=2
(d) Chaindepth=6
4 6
1
3
2 7
5 2 7
1
4
3 5
6
4
6
1
3
2
7
5
h <= w* log n
![Page 64: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/64.jpg)
Complexity of AND/OR Tree Search
AND/OR tree OR tree
Space O(n) O(n)
Time
O(n dh)O(n dw* log n)
(Freuder & Quinn85), (Collin, Dechter & Katz91), (Bayardo & Miranker95), (Darwiche01)
O(dn)
d = domain sizeh = depth of pseudo-treen = number of variablesw*= treewidth
![Page 65: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/65.jpg)
From AND/OR TreeA
D
B C
E
F
A
D
B
CE
F
G H
J
K
G
H
J
KAOR
0AND 1
BOR B
0AND 1 0 1
EOR C E C E C E C
OR D F D F D F D F D F D F D F D F
AND
AND 0 10 1 0 10 1 0 10 1 0 10 1
OR
OR
AND
AND
0
G
H H
0101
0 1
1
G
H H
0101
0 1
0
J
K K
0101
0 1
1
J
K K
0101
0 1
0
G
H H
0101
0 1
1
G
H H
0101
0 1
0
J
K K
0101
0 1
1
J
K K
0101
0 1
0
G
H H
0101
0 1
1
G
H H
0101
0 1
0
J
K K
0101
0 1
1
J
K K
0101
0 1
0
G
H H
0101
0 1
1
G
H H
0101
0 1
0
J
K K
0101
0 1
1
J
K K
0101
0 1
0
G
H H
0101
0 1
1
G
H H
0101
0 1
0
J
K K
0101
0 1
1
J
K K
0101
0 1
0
G
H H
0101
0 1
1
G
H H
0101
0 1
0
J
K K
0101
0 1
1
J
K K
0101
0 1
0
G
H H
0101
0 1
1
G
H H
0101
0 1
0
J
K K
0101
0 1
1
J
K K
0101
0 1
0
G
H H
0101
0 1
1
G
H H
0101
0 1
0
J
K K
0101
0 1
1
J
K K
0101
0 1
![Page 66: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/66.jpg)
An AND/OR GraphA
D
B C
E
F
A
D
B
CE
F
G H
J
K
G
H
J
KAOR
0AND 1
BOR B
0AND 1 0 1
EOR C E C E C E C
OR D F D F D F D F D F D F D F D F
AND
AND 0 10 1 0 10 1 0 10 1 0 10 1
OR
OR
AND
AND
0
G
H H
0101
0 1
1
G
H H
0101
0 1
0
J
K K
0101
0 1
1
J
K K
0101
0 1
![Page 67: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/67.jpg)
Complexity of AND/OR Graph Search
AND/OR graph OR graph
Space O(n dw*) O(n dpw*)
Time O(n dw*) O(n dpw*)
d = domain sizen = number of variablesw*= treewidthpw*= pathwidth
w* ≤ pw* ≤ w* log n
![Page 68: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/68.jpg)
All Four Search Spaces
Full OR search tree
126 nodes
Full AND/OR search tree
54 AND nodes
Context minimal OR search graph
28 nodes
Context minimal AND/OR search graph
18 AND nodes
0 1 0 1 0 1 0 1
0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1
0 1 0 1 0 1 0 10 1 0 1 0 1 0 1 0 1 0 1 0 1 0 10 1 0 1 0 1 0 1 0 1 0 1 0 1 0 10 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1
0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1
0 1 0 1
CD
FE
BA 0 1
AOR
0AND
BOR
0AND
OR E
OR F F
AND 0 1 0 1
AND 0 1
C
D D
0 1 0 1
0 1
1
E
F F
0 1 0 1
0 1
C
D D
0 1 0 1
0 1
1
B
0
E
F F
0 1 0 1
0 1
C
D D
0 1 0 1
0 1
1
E
F F
0 1 0 1
0 1
C
D D
0 1 0 1
0 1
0 1 0 1 0 1 0 1
0 1 0 1 0 1 0 1
0 1
0 1 0 1
0 1 0 1
CD
FE
BA 0 1
AOR0ANDBOR
0ANDOR E
OR F FAND 0 1
AND 0 1
C
D D0 1
0 1
1
EC
D D
0 1
1
B
0
E
F F
0 1
C
1
EC
A
E
C
B
F
D
![Page 69: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/69.jpg)
How Big Is The Context?
Theorem: The maximum context size for a pseudo tree is equal to the treewidth of the graph along the pseudo tree.
C
HK
D
M
F
G
A
B
E
J
O
L
N
P
[AB]
[AF][CHAE]
[CEJ]
[CD]
[CHAB]
[CHA]
[CH]
[C]
[ ]
[CKO]
[CKLN]
[CKL]
[CK]
[C]
(C K H A B E J L N O D P M F G)
B A
C
E
F G
H
J
D
K M
L
N
OP
max context size = treewidth
![Page 70: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/70.jpg)
AND/OR Context Minimal Graph
C
0
K0
H
0
L0 1
N N
0 1 0 1
F F F
1 1
0 1 0 1
F
G
0 1
1
A0 1
B B
0 1 0 1
E EE E
0 1 0 1
J JJ J
0 1 0 1
A0 1
B B
0 1 0 1
E EE E
0 1 0 1
J JJ J
0 1 0 1
G
0 1
G
0 1
G
0 1
M
0 1
M
0 1
M
0 1
M
0 1
P
0 1
P
0 1
O
0 1
O
0 1
O
0 1
O
0 1
L0 1
N N
0 1 0 1
P
0 1
P
0 1
O
0 1
O
0 1
O
0 1
O
0 1
D
0 1
D
0 1
D
0 1
D
0 1
K0
H
0
L0 1
N N
0 1 0 1
1 1
A0 1
B B
0 1 0 1
E EE E
0 1 0 1
J JJ J
0 1 0 1
A0 1
B B
0 1 0 1
E EE E
0 1 0 1
J JJ J
0 1 0 1
P
0 1
P
0 1
O
0 1
O
0 1
O
0 1
O
0 1
L0 1
N N
0 1 0 1
P
0 1
P
0 1
O
0 1
O
0 1
O
0 1
O
0 1
D
0 1
D
0 1
D
0 1
D
0 1
B A
C
E
F G
HJ
D
K M
L
N
OP
C
HK
D
M
F
G
A
B
E
J
O
L
N
P
[AB]
[AF][CHAE]
[CEJ]
[CD]
[CHAB]
[CHA]
[CH]
[C]
[ ]
[CKO]
[CKLN]
[CKL]
[CK]
[C]
(C K H A B E J L N O D P M F G)
Variable Elimination
AND/OR Search
![Page 71: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/71.jpg)
The impact of the pseudo-tree C
0
K0
H0
L0 1
N N0 1 0 1
F F F
1 1
0 1 0 1F
G01
1
A0 1
B B0 1 0 1
E EE E0 1 0 1
J JJ J0 1 0 1
A0 1
B B0 1 0 1
E EE E0 1 0 1
J JJ J0 1 0 1
G01
G01
G01
M01
M01
M01
M01
P01
P01
O01
O01
O01
O01
L0 1
N N0 1 0 1
P01
P01
O01
O01
O01
O01
D01
D01
D01
D01
K0
H0
L0 1
N N0 1 0 1
1 1
A0 1
B B0 1 0 1
E EE E0 1 0 1
J JJ J0 1 0 1
A0 1
B B0 1 0 1
E EE E0 1 0 1
J JJ J0 1 0 1
P01
P01
O01
O01
O01
O01
L0 1
N N0 1 0 1
P01
P01
O01
O01
O01
O01
D01
D01
D01
D01
B A
C
E
F G
HJ
D
K M
L
N
OP
C
HK
D
M
F
G
A
B
E
J
O
L
N
P
[AB]
[AF][CHAE]
[CEJ]
[CD]
[CHAB]
[CHA]
[CH]
[C]
[ ]
[CKO]
[CKLN]
[CKL]
[CK]
[C]
(C K H A B E J L N O D P M F G)
(C D K B A O M L N P J H E F G)
F [AB]
G [AF]
J [ABCD]
D [C]
M [CD]
E [ABCDJ]
B [CD]
A [BCD]
H [ABCJ]
C [ ]
P [CKO]
O [CK]
N [KLO]
L [CKO]
K [C]
C
0 1
D01
N01
D01
K01
K01
B01
B01
B01
B01
M01
M01
M01
M01
A01
A01
A01
A01
A01
A01
A01
A01
F01
F01
F01
F01
G01
G01
G01
G01
J01
J01
J01
J01
J01
J01
J01
J01
J01
J01
J01
J01
J01
J01
J01
J01
E1
E01
E01
E01
E01
E01
E01
E01
E01
E01
E01
E01
E01
E01
E01
E01
E01
E01
E01
E01
E01
E01
E01
E01
E01
E01
E01
E01
E01
E01
E01
E01
H01
H01
H01
H01
H01
H01
H01
H01
H01
H01
H01
H01
H01
H01
H01
H01
O01
O01
O01
O01
L01
L01
L01
L01
L01
L01
L01
L01
P01
P01
P01
P01
P01
P01
P01
P01
N01
N01
N01
N01
N01
N01
N01
What is a good pseudo-tree?How to find a good one?
W=4,h=8
W=5,h=6
![Page 72: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/72.jpg)
Outline
What are graphical models Inference: Exact and approximate Conditioning Search: exact and approximate Hybrids of search and inference (exact) Compilation, (e.g., AND/OR Decision Diagrams) Questions:
Representation issues: directed vs undirected The role of hidden variables Finding good structure Representation guided by human representation Computation: inspired by human thinking
![Page 73: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/73.jpg)
AOBDD vs. OBDD (Mateescu and Dechter 2006)
D
C
B
F
A
E
G
H
1 0
B
CC C
D D D D D
E E
F F F
G G G G
H
AOBDD
18 nonterminals
47 arcs
OBDD
27 nonterminals
54 arcs
A0 1
B0 1
C0 1
0
D0 1
1
F0 1
G0 1
H0 1
C0 1
D0 1
E0 1
F0 1
G0 1
B0 1
C0 1
F0 1
C0 1
F0 1
H0 1
A
B
C F
D E G H
A
B
C
F
D
E
G
H
D
C B F
AE
G
H
primal graph
The context-minimal graphCan be “minimized into an AOMDDBy merging and redundancy removal
![Page 74: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/74.jpg)
Is it consistent?
Find solution NP-complete
Count solutions #P-complete
unminimal const
Always consistent
Find t s.t P(t)>0 Easy: backtrack-free
Find P(X|e)? #P-complete
Explicit minimal tables
Solved by search
Hard to sample
Solved by variable elimination
Easy to sample
Constraint Network vsBayesian Network
Constraint networks Probability networks
CBD
C
D
E
A
B
F
),|( CBDP
),,,,,(represents
FEDCBAPF)E,D,C,B,(A, representssol
C
D
E
A
B
F
![Page 75: Problem Solving with Graphical Modelsdechter/talks/dagshtul-public.pdf · Problem Solving with Graphical Models Rina Dechter Donald Bren School of Computer Science University of California,](https://reader033.fdocuments.net/reader033/viewer/2022060207/5f03f4777e708231d40b986c/html5/thumbnails/75.jpg)
The End
Thank You