algorithms for reasoning with graphical models › ~dechter › courses › ics-276 › spring-18...
TRANSCRIPT
![Page 1: Algorithms for reasoning with graphical models › ~dechter › courses › ics-276 › spring-18 › slid… · Algorithms for reasoning with graphical models Class 1 Rina Dechter](https://reader033.vdocuments.net/reader033/viewer/2022052923/5f0428b27e708231d40c9a42/html5/thumbnails/1.jpg)
Algorithms for reasoning with graphical models
Class 1Rina Dechter
class1 276-2018
Dechter-Morgan&claypool book (Dbook): Chapters 1-2
![Page 2: Algorithms for reasoning with graphical models › ~dechter › courses › ics-276 › spring-18 › slid… · Algorithms for reasoning with graphical models Class 1 Rina Dechter](https://reader033.vdocuments.net/reader033/viewer/2022052923/5f0428b27e708231d40c9a42/html5/thumbnails/2.jpg)
Outline • Graphical models: The constraint network, Probabilistic networks, cost
networks and mixed networks. queries: consistency, counting, optimization and
likelihood queries.
• Inference: Bucket elimination for deterministic networks (Adaptive-
consistency, and the Davis-Putnam algorithms.) The induced-width
• Inference: Bucket-elimination for Bayesian and Markov networks
queries (mpe, map, marginal and probability of evidence)
• Graph properties: induced-width, tree-width, chordal graphs,
hypertrees, join-trees.
• Inference: Tree-decomposition algorithms (join-tree propagation and
junction-trees )
• Approximation by bounded Inference: (weighted Mini-bucket ,
belief/constraint-propagation, constraint propagation, generalized belief
propagation, variational methods)
• Search for csps: Backtracking; pruning search by constraint propagation,
backjumping and learning.
• Search: AND/OR search Spaces for likelihood, optimization queries (Probability
of evidence, Partition function, MAP and MPE queries, AND/OR branch and
bound).
• Approximation by sampling: Gibbs sampling, Importance sampling, cutset-
sampling, SampleSearch and AND/OR sampling, Stochastic Local Search.
• Hybrid of search Inference: cutset-conditioning and cutset-samplingclass1 276-2018
![Page 3: Algorithms for reasoning with graphical models › ~dechter › courses › ics-276 › spring-18 › slid… · Algorithms for reasoning with graphical models Class 1 Rina Dechter](https://reader033.vdocuments.net/reader033/viewer/2022052923/5f0428b27e708231d40c9a42/html5/thumbnails/3.jpg)
Outline
• Graphical models: The constraint network, Probabilistic networks, cost
networks and mixed networks. Graphical representations and queries:
consistency, counting, optimization and likelihood queries.
• Constraints inference: Bucket elimination for deterministic networks
(Adaptive-consistency, and the Davis-Putnam algorithms.) The induced-
width.
• Inference: Bucket-elimination for Bayesian and Markov networks queries (mpe,map,
marginal and probability of evidence)
• Graph properties: induced-width, tree-width, chordal graphs, hypertrees, join-trees.
• Inference: Tree-decomposition algorithms (join-tree propagation and junction-trees
algorithm, Cluster tree-elimination. )
• Approximation by bounded Inference: (Mini-bucket , belief-propagation, constraint
propagation, generalized belief propagation)
• Search: Backtracking search algorithms; pruning search by constraint propagation,
backjumping and learning. class1 276-2018
![Page 4: Algorithms for reasoning with graphical models › ~dechter › courses › ics-276 › spring-18 › slid… · Algorithms for reasoning with graphical models Class 1 Rina Dechter](https://reader033.vdocuments.net/reader033/viewer/2022052923/5f0428b27e708231d40c9a42/html5/thumbnails/4.jpg)
Course Requirements/Textbook
• Homeworks : There will be 5-6 problem sets , graded 70% of the
final grades.
• A term project: paper presentation, a programming project.
• Books:
• “Reasoning with probabilistic and deterministic graphical
models”, R. Dechter, Claypool, 2013
https://www.morganclaypool.com/doi/abs/10.2200/S00529ED1V
01Y201308AIM023
o “Modeling and Reasoning with Bayesian Networks”, A.
Darwiche, MIT Press, 2009.
o “Constraint Processing” , R. Dechter, Morgan Kauffman, 2003
class1 276-2018
![Page 5: Algorithms for reasoning with graphical models › ~dechter › courses › ics-276 › spring-18 › slid… · Algorithms for reasoning with graphical models Class 1 Rina Dechter](https://reader033.vdocuments.net/reader033/viewer/2022052923/5f0428b27e708231d40c9a42/html5/thumbnails/5.jpg)
Outline of classes• Part 1: Introduction and Inference
• Part 2: Search
• Parr 3: Variational Methods and Monte-Carlo Sampling
class1 276-2018
E K
F
LH
C
BA
M
G
J
DABC
BDEF
DGF
EFH
FHK
HJ KLM
0 1 0 1 0 1 0 1
0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1
0101010101010101010101010101010101010101010101010101010101010101
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
EC
FD
BA 0 1
Context minimal AND/OR search graph
AOR0ANDBOR
0ANDOR E
OR F FAND 01
AND 0 1C
D D01
0 1
1EC
D D0 1
1B
0E
F F0 1
C1
EC
![Page 6: Algorithms for reasoning with graphical models › ~dechter › courses › ics-276 › spring-18 › slid… · Algorithms for reasoning with graphical models Class 1 Rina Dechter](https://reader033.vdocuments.net/reader033/viewer/2022052923/5f0428b27e708231d40c9a42/html5/thumbnails/6.jpg)
• Basics of graphical models
– Queries
– Examples, applications, and tasks
– Algorithms overview
• Inference algorithms, exact
– Bucket elimination for trees
– Bucket elimination
– Jointree clustering
– Elimination orders
• Approximate elimination– Decomposition bounds
– Mini-bucket & weighted mini-bucket
– Belief propagation
• Summary and Part 2
class1 276-2018
RoadMap: Introduction and Inference
ABC
BDEF
DGF
EFH
FHK
HJ KLM
A
D E
CBB C
ED
E K
F
L
H
C
BA
M
G
J
D
![Page 7: Algorithms for reasoning with graphical models › ~dechter › courses › ics-276 › spring-18 › slid… · Algorithms for reasoning with graphical models Class 1 Rina Dechter](https://reader033.vdocuments.net/reader033/viewer/2022052923/5f0428b27e708231d40c9a42/html5/thumbnails/7.jpg)
• Basics of graphical models
– Queries
– Examples, applications, and tasks
– Algorithms overview
• Inference algorithms, exact
– Bucket elimination for trees
– Bucket elimination
– Jointree clustering
– Elimination orders
• Approximate elimination
– Decomposition bounds
– Mini-bucket & weighted mini-bucket
– Belief propagation
• Summary and Class 2
class1 276-2018
RoadMap: Introduction and Inference
ABC
BDEF
DGF
EFH
FHK
HJ KLM
A
D E
CBB C
ED
E K
F
L
H
C
BA
M
G
J
D
![Page 8: Algorithms for reasoning with graphical models › ~dechter › courses › ics-276 › spring-18 › slid… · Algorithms for reasoning with graphical models Class 1 Rina Dechter](https://reader033.vdocuments.net/reader033/viewer/2022052923/5f0428b27e708231d40c9a42/html5/thumbnails/8.jpg)
Probabilistic Graphical models• Describe structure in large problems
– Large complex system
– Made of “smaller”, “local” interactions
– Complexity emerges through interdependence
class1 276-2018
![Page 9: Algorithms for reasoning with graphical models › ~dechter › courses › ics-276 › spring-18 › slid… · Algorithms for reasoning with graphical models Class 1 Rina Dechter](https://reader033.vdocuments.net/reader033/viewer/2022052923/5f0428b27e708231d40c9a42/html5/thumbnails/9.jpg)
Probabilistic Graphical models• Describe structure in large problems
– Large complex system
– Made of “smaller”, “local” interactions
– Complexity emerges through interdependence
• Examples & Tasks– Maximization (MAP): compute the most probable configuration
[Yanover & Weiss 2002]
class1 276-2018
![Page 10: Algorithms for reasoning with graphical models › ~dechter › courses › ics-276 › spring-18 › slid… · Algorithms for reasoning with graphical models Class 1 Rina Dechter](https://reader033.vdocuments.net/reader033/viewer/2022052923/5f0428b27e708231d40c9a42/html5/thumbnails/10.jpg)
Probabilistic Graphical models• Describe structure in large problems
– Large complex system
– Made of “smaller”, “local” interactions
– Complexity emerges through interdependence
• Examples & Tasks– Summation & marginalization
grass
plane
sky
grass
cow
Observation y Observation yMarginals p( xi | y ) Marginals p( xi | y )
and
“partition function”
class1 276-2018
e.g., [Plath et al. 2009]
![Page 11: Algorithms for reasoning with graphical models › ~dechter › courses › ics-276 › spring-18 › slid… · Algorithms for reasoning with graphical models Class 1 Rina Dechter](https://reader033.vdocuments.net/reader033/viewer/2022052923/5f0428b27e708231d40c9a42/html5/thumbnails/11.jpg)
Graphical models• Describe structure in large problems
– Large complex system
– Made of “smaller”, “local” interactions
– Complexity emerges through interdependence
• Examples & Tasks– Mixed inference (marginal MAP, MEU, …)
Test
DrillOil sale
policy
Test
result
Seismic
structure
Oil
underground
Oil
produced
Test
cost
Drill
cost
Sales
cost
Oil
sales
Market
information
Influence diagrams &optimal decision-making
(the “oil wildcatter” problem)
class1 276-2018
e.g., [Raiffa 1968; Shachter 1986]
![Page 12: Algorithms for reasoning with graphical models › ~dechter › courses › ics-276 › spring-18 › slid… · Algorithms for reasoning with graphical models Class 1 Rina Dechter](https://reader033.vdocuments.net/reader033/viewer/2022052923/5f0428b27e708231d40c9a42/html5/thumbnails/12.jpg)
class1 276-2018
In more details…
![Page 13: Algorithms for reasoning with graphical models › ~dechter › courses › ics-276 › spring-18 › slid… · Algorithms for reasoning with graphical models Class 1 Rina Dechter](https://reader033.vdocuments.net/reader033/viewer/2022052923/5f0428b27e708231d40c9a42/html5/thumbnails/13.jpg)
A Bred green
red yellow
green red
green yellow
yellow green
yellow red
Example: map coloring
Variables - countries (A,B,C,etc.)
Values - colors (red, green, blue)
Constraints: etc. ,ED D, AB,A
C
A
B
D
E
F
G
Constraint Networks
A
B
E
G
DF
C
Constraint graph
class1 276-2018
![Page 14: Algorithms for reasoning with graphical models › ~dechter › courses › ics-276 › spring-18 › slid… · Algorithms for reasoning with graphical models Class 1 Rina Dechter](https://reader033.vdocuments.net/reader033/viewer/2022052923/5f0428b27e708231d40c9a42/html5/thumbnails/14.jpg)
Propositional Reasoning
• If Alex goes, then Becky goes:
• If Chris goes, then Alex goes:
• Question:Is it possible that Chris goes to the party but Becky does not?
Example: party problem
BA
A C
e?satisfiabl ,,
theIs
C B, ACBA
theorynalpropositio
AB
C
class1 276-2018
![Page 15: Algorithms for reasoning with graphical models › ~dechter › courses › ics-276 › spring-18 › slid… · Algorithms for reasoning with graphical models Class 1 Rina Dechter](https://reader033.vdocuments.net/reader033/viewer/2022052923/5f0428b27e708231d40c9a42/html5/thumbnails/15.jpg)
Bayesian Networks (Pearl 1988)
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
Dyspnoea
P(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
• Posterior marginals, probability of evidence, MPE
• P( D= 0) = σ𝑆,𝐿,𝐵,𝑋 P(S)· P(C|S)· P(B|S)· P(X|C,S)· P(D|C,B
MAP(P)= 𝑚𝑎𝑥𝑆,𝐿,𝐵,𝑋 P(S)· P(C|S)· P(B|S)· P(X|C,S)· P(D|C,B)
Combination: ProductMarginalization: sum/max
class1 276-2018
![Page 16: Algorithms for reasoning with graphical models › ~dechter › courses › ics-276 › spring-18 › slid… · Algorithms for reasoning with graphical models Class 1 Rina Dechter](https://reader033.vdocuments.net/reader033/viewer/2022052923/5f0428b27e708231d40c9a42/html5/thumbnails/16.jpg)
Probabilistic reasoning (directed)
• Alex is-likely-to-go in bad weather• Chris rarely-goes in bad weather• Becky is indifferent but unpredictable
Questions:• Given bad weather, which group of individuals is most
likely to show up at the party? • What is the probability that Chris goes to the party
but Becky does not?
Party example: the weather effect
P(W,A,C,B) = P(B|W) · P(C|W) · P(A|W) · P(W)
P(A,C,B|W=bad) = 0.9 · 0.1 · 0.5
P(A|W=bad)=.9W A
P(C|W=bad)=.1W C
P(B|W=bad)=.5W B
W
P(W)
P(A|W)
P(C|W)P(B|W)
B CA
W A P(A|W)
good 0 .01
good 1 .99
bad 0 .1
bad 1 .9
class1 276-2018
![Page 17: Algorithms for reasoning with graphical models › ~dechter › courses › ics-276 › spring-18 › slid… · Algorithms for reasoning with graphical models Class 1 Rina Dechter](https://reader033.vdocuments.net/reader033/viewer/2022052923/5f0428b27e708231d40c9a42/html5/thumbnails/17.jpg)
Alarm network• Bayes nets: compact representation of large joint distributions
PCWP CO
HRBP
HREKG HRSAT
ERRCAUTERHRHISTORY
CATECHOL
SAO2 EXPCO2
ARTCO2
VENTALV
VENTLUNG VENITUBE
DISCONNECT
MINVOLSET
VENTMACHKINKEDTUBEINTUBATIONPULMEMBOLUS
PAP SHUNT
ANAPHYLAXIS
MINOVL
PVSAT
FIO2
PRESS
INSUFFANESTHTPR
LVFAILURE
ERRBLOWOUTPUTSTROEVOLUMELVEDVOLUME
HYPOVOLEMIA
CVP
BP
The “alarm” network: 37 variables, 509 parameters (rather than 237 = 1011 !)
[Beinlich et al., 1989]
class1 276-2018
![Page 18: Algorithms for reasoning with graphical models › ~dechter › courses › ics-276 › spring-18 › slid… · Algorithms for reasoning with graphical models Class 1 Rina Dechter](https://reader033.vdocuments.net/reader033/viewer/2022052923/5f0428b27e708231d40c9a42/html5/thumbnails/18.jpg)
18
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
),,|,( ACBAbadwBCP
A→B C→A
B A CP(C|W)P(B|W)
P(W)
P(A|W)
W
B A C
A→B C→A
B A C
Alex is-likely-to-go in bad weatherChris rarely-goes in bad weatherBecky is indifferent but unpredictable
![Page 19: Algorithms for reasoning with graphical models › ~dechter › courses › ics-276 › spring-18 › slid… · Algorithms for reasoning with graphical models Class 1 Rina Dechter](https://reader033.vdocuments.net/reader033/viewer/2022052923/5f0428b27e708231d40c9a42/html5/thumbnails/19.jpg)
Graphical models (cost networks)Example:
The combination operator defines an overall function from the individual factors,
e.g., “+” :
Notation:
Discrete Xi values called states
Tuple or configuration: states taken by a set of variables
Scope of f: set of variables that are arguments to a factor f
often index factors by their scope, e.g.,
class1 276-2018
A graphical model consists of:
-- variables
-- domains
-- functions or “factors”
and a combination operator
(we’ll assume discrete)
![Page 20: Algorithms for reasoning with graphical models › ~dechter › courses › ics-276 › spring-18 › slid… · Algorithms for reasoning with graphical models Class 1 Rina Dechter](https://reader033.vdocuments.net/reader033/viewer/2022052923/5f0428b27e708231d40c9a42/html5/thumbnails/20.jpg)
Graphical models (cost networks)
+= 0 + 6
A B f(A,B)
0 0 6
0 1 0
1 0 0
1 1 6
B C f(B,C)
0 0 6
0 1 0
1 0 0
1 1 6
A B C f(A,B,C)
0 0 0 12
0 0 1 6
0 1 0 0
0 1 1 6
1 0 0 6
1 0 1 0
1 1 0 6
1 1 1 12
=
For discrete variables, think of functions as “tables”
(though we might represent them more efficiently)
A graphical model consists of:
-- variables
-- domains
-- functions or “factors”
and a combination operator
Example:
(we’ll assume discrete)
class1 276-2018
![Page 21: Algorithms for reasoning with graphical models › ~dechter › courses › ics-276 › spring-18 › slid… · Algorithms for reasoning with graphical models Class 1 Rina Dechter](https://reader033.vdocuments.net/reader033/viewer/2022052923/5f0428b27e708231d40c9a42/html5/thumbnails/21.jpg)
Graph Visualiization: Primal Graph
Primal graph:variables → nodesfactors → cliques
G
A
B C
D F
A graphical model consists of:
-- variables
-- domains
-- functions or “factors”
and a combination operator
class1 276-2018
![Page 22: Algorithms for reasoning with graphical models › ~dechter › courses › ics-276 › spring-18 › slid… · Algorithms for reasoning with graphical models Class 1 Rina Dechter](https://reader033.vdocuments.net/reader033/viewer/2022052923/5f0428b27e708231d40c9a42/html5/thumbnails/22.jpg)
Example: Constraint networks
Overall function is “and” of individual constraints:
for adjacent regions i,j
“Tabular” form: X0 X1 f(X0 ,X1)
0 0 0
0 1 1
0 2 1
1 0 1
1 1 0
1 2 1
2 0 1
2 1 1
2 2 0Tasks: “max”: is there a solution?
“sum”: how many solutions?
class1 276-2018
![Page 23: Algorithms for reasoning with graphical models › ~dechter › courses › ics-276 › spring-18 › slid… · Algorithms for reasoning with graphical models Class 1 Rina Dechter](https://reader033.vdocuments.net/reader033/viewer/2022052923/5f0428b27e708231d40c9a42/html5/thumbnails/23.jpg)
habits. smoking similar have Friends
cancer. causes Smoking
Markov logic, Markov networks
)()(),(,
)()(
ySmokesxSmokesyxFriendsyx
xCancerxSmokesx
1.1
5.1
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)
SA CA f(SA,CA)
0 0 exp(1.5)
0 1 exp(1.5)
1 0 1.0
1 1 exp(1.5)
FAB SA SB f(.)
0 0 0 exp(1.1)
0 0 1 exp(1.1)
0 1 0 exp(1.1)
0 1 1 exp(1.1)
1 0 0 exp(1.1)
1 0 1 1.0
1 1 0 1.0
1 1 1 exp(1.1)
[Richardson & Domingos 2005]
class1 276-2018
![Page 24: Algorithms for reasoning with graphical models › ~dechter › courses › ics-276 › spring-18 › slid… · Algorithms for reasoning with graphical models Class 1 Rina Dechter](https://reader033.vdocuments.net/reader033/viewer/2022052923/5f0428b27e708231d40c9a42/html5/thumbnails/24.jpg)
Graphical visualization
Primal graph:variables nodesfactors cliques
A graphical model consists of:
-- variables
-- domains
-- functions or “factors”
and a combination operator
class1 276-2018
G
A
B C
D F
ABD
BCF AC
DFGD
B
C
A F
Dual graph:factor scopes nodesedges intersections (separators)
![Page 25: Algorithms for reasoning with graphical models › ~dechter › courses › ics-276 › spring-18 › slid… · Algorithms for reasoning with graphical models Class 1 Rina Dechter](https://reader033.vdocuments.net/reader033/viewer/2022052923/5f0428b27e708231d40c9a42/html5/thumbnails/25.jpg)
Graphical visualization“Factor” graph: explicitly indicate the scope of each factor
variables circlesfactors squares
G
A
B C
D F
A
B
C
D
A
B
C
D
Useful for disambiguating factorization:
= vs.
O(d4) pairwise: O(d2)
A
B
C
D
?
class1 276-2018
![Page 26: Algorithms for reasoning with graphical models › ~dechter › courses › ics-276 › spring-18 › slid… · Algorithms for reasoning with graphical models Class 1 Rina Dechter](https://reader033.vdocuments.net/reader033/viewer/2022052923/5f0428b27e708231d40c9a42/html5/thumbnails/26.jpg)
Graphical modelsA graphical model consists of:
-- variables
-- domains
-- functions or “factors”
Operators:
combination operator
(sum, product, join, …)
elimination operator
(projection, sum, max, min, ...)
Types of queries:
Marginal:
MPE / MAP:
Marginal MAP:
class1 276-2018
)( : CAFfi
A
D
BC
E
F
• All these tasks are NP-hard
• exploit problem structure
• identify special cases
• approximate
A C F P(F|A,C)
0 0 0 0.14
0 0 1 0.96
0 1 0 0.40
0 1 1 0.60
1 0 0 0.35
1 0 1 0.65
1 1 0 0.72
1 1 1 0.68
Conditional Probability Table (CPT)
Primal graph(interaction graph)
A C F
red green blue
blue red red
blue blue green
green red blue
Relation
(𝐴⋁𝐶⋁𝐹)
![Page 27: Algorithms for reasoning with graphical models › ~dechter › courses › ics-276 › spring-18 › slid… · Algorithms for reasoning with graphical models Class 1 Rina Dechter](https://reader033.vdocuments.net/reader033/viewer/2022052923/5f0428b27e708231d40c9a42/html5/thumbnails/27.jpg)
Graphical models/reasoning task
class1 276-2018
![Page 28: Algorithms for reasoning with graphical models › ~dechter › courses › ics-276 › spring-18 › slid… · Algorithms for reasoning with graphical models Class 1 Rina Dechter](https://reader033.vdocuments.net/reader033/viewer/2022052923/5f0428b27e708231d40c9a42/html5/thumbnails/28.jpg)
Summary of graphical models types
• Constraint networks
• Cost networks
• Bayesian network
• Markov networks
• Mixed probability and constraint network
• Influence diagrams
class1 276-2018
![Page 29: Algorithms for reasoning with graphical models › ~dechter › courses › ics-276 › spring-18 › slid… · Algorithms for reasoning with graphical models Class 1 Rina Dechter](https://reader033.vdocuments.net/reader033/viewer/2022052923/5f0428b27e708231d40c9a42/html5/thumbnails/29.jpg)
30
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
D
E
F
G
Constraint Networks
Constraint graph
A
BD
CG
F
E
Queries: Find one solution, all solutions, countingclass1 276-2018
Combination = joinMarginalization = projection
![Page 30: Algorithms for reasoning with graphical models › ~dechter › courses › ics-276 › spring-18 › slid… · Algorithms for reasoning with graphical models Class 1 Rina Dechter](https://reader033.vdocuments.net/reader033/viewer/2022052923/5f0428b27e708231d40c9a42/html5/thumbnails/30.jpg)
Example of a Cost Network
Combination: sumMarginalization:min/max
class1 276-2018
![Page 31: Algorithms for reasoning with graphical models › ~dechter › courses › ics-276 › spring-18 › slid… · Algorithms for reasoning with graphical models Class 1 Rina Dechter](https://reader033.vdocuments.net/reader033/viewer/2022052923/5f0428b27e708231d40c9a42/html5/thumbnails/31.jpg)
A Bayesian Network
class1 276-2018
Combination: productMarginalization: sum or min/max
![Page 32: Algorithms for reasoning with graphical models › ~dechter › courses › ics-276 › spring-18 › slid… · Algorithms for reasoning with graphical models Class 1 Rina Dechter](https://reader033.vdocuments.net/reader033/viewer/2022052923/5f0428b27e708231d40c9a42/html5/thumbnails/32.jpg)
Markov Networks
set2 huji
![Page 33: Algorithms for reasoning with graphical models › ~dechter › courses › ics-276 › spring-18 › slid… · Algorithms for reasoning with graphical models Class 1 Rina Dechter](https://reader033.vdocuments.net/reader033/viewer/2022052923/5f0428b27e708231d40c9a42/html5/thumbnails/33.jpg)
Example domains for graphical models• Natural Language processing
– Information extraction, semantic parsing, translation, topic models, …
• Computer vision– Object recognition, scene analysis, segmentation, tracking, …
• Computational biology– Pedigree analysis, protein folding and binding, sequence matching, …
• Networks– Webpage link analysis, social networks, communications, citations, ….
• Robotics– Planning & decision making
class1 276-2018
![Page 34: Algorithms for reasoning with graphical models › ~dechter › courses › ics-276 › spring-18 › slid… · Algorithms for reasoning with graphical models Class 1 Rina Dechter](https://reader033.vdocuments.net/reader033/viewer/2022052923/5f0428b27e708231d40c9a42/html5/thumbnails/34.jpg)
Complexity of Reasoning Tasks
• Constraint satisfaction
• Counting solutions
• Combinatorial optimization
• Belief updating
• Most probable explanation
• Decision-theoretic planning
0
200
400
600
800
1000
1200
1 2 3 4 5 6 7 8 9 10
f(n)
n
Linear / Polynomial / Exponential
Linear
Polynomial
Exponential
Reasoning iscomputationally hard
Complexity is
Time and space(memory)class1 276-2018
![Page 35: Algorithms for reasoning with graphical models › ~dechter › courses › ics-276 › spring-18 › slid… · Algorithms for reasoning with graphical models Class 1 Rina Dechter](https://reader033.vdocuments.net/reader033/viewer/2022052923/5f0428b27e708231d40c9a42/html5/thumbnails/35.jpg)
• Basics of graphical models
– Queries
– Examples, applications, and tasks
– Algorithms overview
• Inference algorithms, exact
– Bucket elimination for trees
– Bucket elimination
– Jointree clustering
– Elimination orders
• Approximate elimination
– Decomposition bounds
– Mini-bucket & weighted mini-bucket
– Belief propagation
• Summary and Class 2
class1 276-2018
RoadMap: Introduction and Inference
ABC
BDEF
DGF
EFH
FHK
HJ KLM
A
D E
CBB C
ED
E K
F
L
H
C
BA
M
G
J
D
![Page 36: Algorithms for reasoning with graphical models › ~dechter › courses › ics-276 › spring-18 › slid… · Algorithms for reasoning with graphical models Class 1 Rina Dechter](https://reader033.vdocuments.net/reader033/viewer/2022052923/5f0428b27e708231d40c9a42/html5/thumbnails/36.jpg)
Sum-Inference
Max-Inference
Mixed-Inference
Types of queries
• NP-hard: exponentially many terms
• We will focus on approximation algorithms
– Anytime: very fast & very approximate ! Slower & more accurate
Hard
er
class1 276-2018
![Page 37: Algorithms for reasoning with graphical models › ~dechter › courses › ics-276 › spring-18 › slid… · Algorithms for reasoning with graphical models Class 1 Rina Dechter](https://reader033.vdocuments.net/reader033/viewer/2022052923/5f0428b27e708231d40c9a42/html5/thumbnails/37.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
class1 276-2018
![Page 38: Algorithms for reasoning with graphical models › ~dechter › courses › ics-276 › spring-18 › slid… · Algorithms for reasoning with graphical models Class 1 Rina Dechter](https://reader033.vdocuments.net/reader033/viewer/2022052923/5f0428b27e708231d40c9a42/html5/thumbnails/38.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.
class1 276-2018
![Page 39: Algorithms for reasoning with graphical models › ~dechter › courses › ics-276 › spring-18 › slid… · Algorithms for reasoning with graphical models Class 1 Rina Dechter](https://reader033.vdocuments.net/reader033/viewer/2022052923/5f0428b27e708231d40c9a42/html5/thumbnails/39.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)
class1 276-2018
![Page 40: Algorithms for reasoning with graphical models › ~dechter › courses › ics-276 › spring-18 › slid… · Algorithms for reasoning with graphical models Class 1 Rina Dechter](https://reader033.vdocuments.net/reader033/viewer/2022052923/5f0428b27e708231d40c9a42/html5/thumbnails/40.jpg)
Conditioning and Cycle cutset
CP
J A
L
B
E
DF M
O
H
K
G N
CP
J
L
B
E
DF M
O
H
K
G N
A
CP
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}
CP
J A
L
B
E
DF M
O
H
K
G N
CP
J
L
B
E
DF M
O
H
K
G N
CP
J
L
E
DF M
O
H
K
G N
CP
J A
L
B
E
DF M
O
H
K
G N
class1 276-2018
![Page 41: Algorithms for reasoning with graphical models › ~dechter › courses › ics-276 › spring-18 › slid… · Algorithms for reasoning with graphical models Class 1 Rina Dechter](https://reader033.vdocuments.net/reader033/viewer/2022052923/5f0428b27e708231d40c9a42/html5/thumbnails/41.jpg)
Search over the Cutset
A=yellow A=green
B=red B=blue B=red B=blueB=green B=yellow
C
K
G
L
D
FH
M
J
E
C
K
G
L
D
FH
M
J
E
C
K
G
L
D
FH
M
J
E
C
K
G
L
D
FH
M
J
E
C
K
G
L
D
FH
M
J
E
C
K
G
L
D
FH
M
J
E
• Inference may require too much memory
• Condition on some of the variables
AC
B K
G
L
D
FH
M
J
E
GraphColoringproblem
class1 276-2018
![Page 42: Algorithms for reasoning with graphical models › ~dechter › courses › ics-276 › spring-18 › slid… · Algorithms for reasoning with graphical models Class 1 Rina Dechter](https://reader033.vdocuments.net/reader033/viewer/2022052923/5f0428b27e708231d40c9a42/html5/thumbnails/42.jpg)
Inference
exp(w*) 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
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 SearchExp(w*) timeO(w*) space
EK
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
FH
M
J
EACB K
G
LD
FH
M
J
E
CK
G
LD
FH
M
J
E
CK
G
LD
FH
M
J
E
CK
G
LD
FH
M
J
E
Search+inference:Space: exp(q)Time: exp(q+c(q))
q: usercontrolled
Bird's-eye View of Exact Algorithms
class1 276-2018
![Page 43: Algorithms for reasoning with graphical models › ~dechter › courses › ics-276 › spring-18 › slid… · Algorithms for reasoning with graphical models Class 1 Rina Dechter](https://reader033.vdocuments.net/reader033/viewer/2022052923/5f0428b27e708231d40c9a42/html5/thumbnails/43.jpg)
Inference
exp(w*) 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
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 SearchExp(w*) timeO(w*) space
EK
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
FH
M
J
EACB K
G
LD
FH
M
J
E
CK
G
LD
FH
M
J
E
CK
G
LD
FH
M
J
E
CK
G
LD
FH
M
J
E
Search+inference:Space: exp(q)Time: exp(q+c(q))
q: usercontrolled
Context minimal AND/OR search graph
18 AND nodes
AOR0ANDBOR
0AND
OR E
OR F F
AND0 1
AND 0 1
C
D D
0 1
0 1
1
EC
D D
0 1
1
B
0
E
F F
0 1
C
1
EC
Bird's-eye View of Exact Algorithms
class1 276-2018
![Page 44: Algorithms for reasoning with graphical models › ~dechter › courses › ics-276 › spring-18 › slid… · Algorithms for reasoning with graphical models Class 1 Rina Dechter](https://reader033.vdocuments.net/reader033/viewer/2022052923/5f0428b27e708231d40c9a42/html5/thumbnails/44.jpg)
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
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
EK
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
FH
M
J
EACB K
G
LD
FH
M
J
E
CK
G
LD
FH
M
J
E
CK
G
LD
FH
M
J
E
CK
G
LD
FH
M
J
E
Inference
Bounded Inference
Search
Sampling
Search + inference:
Sampling + bounded inference
Bird's-eye View of Approximate Algorithms
class1 276-2018
Context minimal AND/OR search graph
18 AND nodes
AOR0ANDBOR
0AND
OR E
OR F F
AND0 1
AND 0 1
C
D D
0 1
0 1
1
EC
D D
0 1
1
B
0
E
F F
0 1
C
1
EC