Date post: | 19-Dec-2015 |
Category: |
Documents |
View: | 214 times |
Download: | 0 times |
1
Bayesian Networks and
Causal Modelling
Ann Nicholson
School of Computer Science and Software Engineering
Monash University
2
Overview
Introduction to Bayesian Networks (BNs) Summary of BN research projects Varieties of Causal intervention
» PRICAI2004: K. Korb, L. Hope, A. Nicholson, K. Axnick
Learning Causal Structure» CaMML software
3
Probability theory for representing uncertainty
Assigns a numerical degree of belief between 0 and 1 to facts» e.g. “it will rain today” is T/F. » P(“it will rain today”) = 0.2 prior probability
(unconditional) Posterior probability (conditional)
» P(“it wil rain today” | “rain is forecast”) = 0.8 Bayes’ Rule: P(H|E) = P(E|H) x P(H) P(E)
4
Bayesian networks
A Bayesian Network (BN) represents a probability distribution graphically (directed acyclic graphs)
Nodes: random variables,» R: “it is raining”, discrete values T/F» T: temperature, cts or discrete variable» C: colour, discrete values {red,blue,green}
Arcs indicate conditional dependencies between variables
P(A,S,T) can be decomposed to P(A)P(S|A)P(T|A)
5
Bayesian networks (cont.)
There is a conditional probability distribution (CPD or CPT) associated with each node.
– probability of each state given parent states
“Jane has the flu”
“Jane has a high temp”
“Thermometertemp reading”
XFlu
YTe
QTh
Models causal relationship
Models possible sensor error
P(Flu=T) = 0.05
P(Te=High|Flu=T) = 0.4P(Te=High|Flu=F) = 0.01
P(Th=High|Te=H) = 0.95P(Th=High|Te=L) = 0.1
6
BN inference
Evidence: observation of specific state Task: compute the posterior probabilities for query
node(s) given evidence.
Th
Y
Flu
Te
Diagnostic inference
Th
Flu
YTe
Predictive inference
Intercausal inference
Te
Flu TBFlu
Mixed inference
Th
Flu
Te
7
Causal Networks
Arcs follow the direction of causal process
Causal Networks are always BNs Bayesian Networks aren't always causal
8
Early BN-related projects
DBNS for discrete monitoring (PhD, 1992) Approximate BN inference algorithms based
on a mutual information measure for relevance (with Nathalie Jitnah, 1996-1999)
Plan recognition: DBNs for predicting users actions and goals in an adventure game (with
David Albrecht, Ingrid Zukerman, 1997-2000) DBNs for ambulation monitoring and fall
diagnosis (with biomedical engineering, 1996-2000) Bayesian Poker (with Kevin Korb, 1996-2003)
9
Knowledge Engineering with BNs
Seabreeze prediction: joint project with Bureau of Meteorology» Comparison of existing simple rule, expert elicited
BN, and BNs from Tetrad-II and CaMML ITS for decimal misconceptions Methodology and tools to support knowledge
engineering process» Matilda: visualisation of d-separation » Support for sensitivity analysis
Written a textbook: » Bayesian Artificial Intelligence, Kevin B. Korb and
Ann E. Nicholson, Chapman & Hall / CRC, 2004.www.csse.monash.edu.au/bai/book
10
Current BN-related projects
BNs for Epidemiology (with Kevin Korb, Charles Twardy)
» ARC Discovery Grant, 2004 » Looking at Coronary Heart Disease data sets» Learning hybrid networks: cts and discrete variables.
BNs for supporting meteorological forecasting process (DSS’2004) (with Ph. D student Tal Boneh, K. Korb, BoM)
» Building domain ontology (in Protege) from expert elicitation» Automatically generating BN fragments» Case studies: Fog, hailstorms, rainfall.
Ecological risk assessment » Goulburn Water, native fish abundance
» Sydney Harbour Water Quality
11
Other projects
Autonomous aircraft monitoring and replanning (with Ph.D. student Tim Wilkin, PRICAI2000,
IAV2004) Dynamic non-uniform abstraction for
approximate planning with MDPs (with Ph.D. student Jiri Baum)
12
Observation and Intervention
Inference from observations» Predictive reasoning (finding effects)» Diagnostic reasoning (finding causes)
Inference with interventions» Predictive reasoning» Not diagnostic reasoning
Causal reasoning shouldn't go against causality. Th
Flu
Te
Diagnostic inference
Th
Flu
Te
Predictive inference
13
Pearlian Determinism
Pearl's reasons for determinism:» Determinism is intuitive» Counterfactuals and causal explanation only make
sense with a deterministic interpretation» Any indeterministic model can be transformed into
a deterministic model
We see no reason for assuming determinism
14
Defining Intervention I
Arc cutting» More intuitive» Intervention node
Intervention node» More general interventions» Much easier to implement» To simulate arc cutting: P(C| c, Ic)=1
Arc cutting isn’t general enough
15
Defining Intervention II
We define an intervention on model M as: M augmented with Ic (M') where:
1. Ic has the purpose of manipulating C
2. Ic is exogenous (has no parents) in M'
3. Ic directly causes (is a parent of) C
To preserve the original network:
PM'(C| c,¬ Ic) = PM' (C| c)
where c are the original parents of C.We also define P*(C) as the intended distribution.
16
Varieties of Intervention: Dependency
The degree of dependency of the effect upon existing parents.
• An independent intervention cuts the child off from its other parents. Thus,
PM'(C| c, Ic) = P*(C)
• A dependent intervention allows any parent interaction.
17
Varieties of Intervention : Indeterminism
The degree of indeterminism of the effect.
A deterministic intervention sets the child to one particular state.
A stochastic intervention sets the child to a positive distribution.
Dependency and Determinism characterize any intervention Pearlian interventions are independent and
deterministic
18
Varieties of Intervention : Effectiveness
We've found the idea of effectiveness useful.
If P*(C) is what's intended and r is the effectiveness, then
PM'(C | c, Ic) = r × P*(C) + (1-r) × PM'(C | c)
This is a dependent intervention.
20
Summary of Causal Intervention
A taxonomy of intervention types More realistic interventions (e.g., partial
effectiveness) A GUI which handles some varieties of
intervention» Pearlian» Partially effective» Extensible to deal with other types of interaction
explicitly
21
Learning Causal Structure
This is the real problem; parameterizing models is relatively straightforward estimation problem.
Size of the dag space is superexponential:» Number of possible orderings: n!» Times number of possible arcs: Cn
2
» Minus number of possible cyclic graphs More exactly (Robinson, 1977):
f(n) = (-1)i+1 Cni 2i(n-i)f(n-i)
so for» n=3, f(n)=25» n=5, f(n)=25,000» n=10, f(n) 4.2x1018
22
Learning Causal Structure
There are two basic methods:» Learning from conditional independencies (CI
learning)» Learning using a scoring metric (Metric learning)
CI learning (Verma and Pearl, 1991)» Suppose you have an Oracle who can answer yes
or no to any question of the type:
is X conditional independence Y given S?» Then you can learn the correct causal model, up to
statistical equivalence (patterns).
23
Statistical Equivalence
Two causal models H1 and H2 are statistically equivalent iff they contain the same variables and joint samples over them provide no statistical grounds for preferring one over the other.
Examples» All fully connected models are equivalent.» A B C and A B C.» A B D C and A B D C.
24
Statistical Equivalence (cont.)
(Verma and Pearl, 1991): Any two causal models over the same variables which have the same skeleton (undirected arcs) and the same directed v-structures are statistically equivalent.
Chickering (1995): If H1 and H2 are statistically equivalent, then they have the same maximum likelihoods relative to any joint samplesmax P(e|H1,1) = max P(e|H2,2)
where i is a parameterization of Hi
25
Other approaches to structure learning
TETRAD II: Spirtes, Glymour and Scheines (1993). Implemented in their PC algorithm» Doesn't handle well with weak links and small samples
(demonstrated empirically in Dai, Korb, Wallace & Wu (1997)).
Bayesian LBN: Cooper & Herskovits' K2 (1991, 1992)» Compute P(hi|e) by brute force, under the various
assumptions which reduce the computation of PCH(h,e) to a polynomial time counting problem.
» But the hypothesis space is exponential; they go for dramatic simplification by assuming we know the temporal ordering of the variables.
26
Learning Variable Order
Reliance upon a given variable order is a major drawback to K2» And many other algorithms (Buntine 1991, Bouckert 1994, Suzuki
1996, Madigan & Raftery 1994)
What's wrong with that?» We want autonomous AI (data mining). If experts can
order the variables they can likely supply models.» Determining variable ordering is half the problem. If
we know A comes before B, the only remaining issue is whether there is a link between the two.
» The number of orderings consistent with dags is exponential (Brightwell & Winkler 1990; number complete). So iterating over all possible orderings will not scale up.
27
Statistical Equivalence Learners
Heckerman & Geiger (1995) advocate learning only up to statistical equivalence classes (a la TETRAD II).» Since observational data cannot distinguish btw
equivalent models, there's no point trying to go further.
Madigan, Andersson, Perlman & Volinsky (1996) follow this advice, use uniform prior over equivalence classes.
Geiger and Heckerman (1994) define Bayesian metrics for linear and discrete equivalence classes of models (BGe and BDe)
28
Statistical Equivalence Learners
Wallace & Korb (1999): This is not right! These are causal models; they are distinguishable on
experimental data.» Failure to collect some data is no reason to change prior
probabilities. E.g., If your thermometer topped out at 35C, you wouldn't
treat 35C and 34C as equally likely. Not all equivalence classes are created equal:{ A B C, A B C, A B C }{ A B C } Within classes some dags should have greater priors
than others… E.g.,» LightsOn InOffice LoggedOn v.» LightsOn InOffice LoggedOn
29
Full Causal Learners
So… a full causal learner is an algorithm that:1. Learns causal connectedness.2. Learns v-structures. Hence, learns equivalence
classes.3. Learns full variable order. Hence, learns full causal
structure (order + connectedness).
TETRAD II: 1, 2. Madigan et al.; Heckerman & Geiger (BGe, BDe): 1,
2. Cooper & Herskovits' K2: 1. Lam and Bacchus MDL: 1, 2 (partial), 3 (partial). Wallace, Neil, Korb MML: 1, 2, 3.
30
CaMML
Minimum Message Length (Wallace \& Boulton 1968) uses Shannon's measure of information:
I(m) = - log P(m) Applied in reverse, we can compute P(h,e) from I(h,e). Given an efficient joint encoding method for the
hypothesis & evidence space (i.e., satisfying Shannon's law), MML:
Searches {hi} for that hypothesis h that minimizes I(h) + I(e|h).
Applies a trade-off between» Model simplicity» Data fit
Equivalent to that h that maximizes P(h)P(e|h) --- i.e., P(h|e).
31
MML search algorithms
MML metrics need to be combined with search. This has been done three ways:
1. Wallace, Korb, Dai (1996): greedy search (linear).» Brute force computation of linear extensions (small models
only)
2. Neil and Korb (1999): genetic algorithms (linear).» Asymptotic estimator of linear extensions» GA chromosomes = causal models» Genetic operators manipulate them» Selection pressure is based on MML
Wallace and Korb (1999): MML sampling (linear, discrete).» Stochastic sampling through space of totally ordered causal
models» No counting of linear extensions required
32
Empirical Results
A weakness in this area --- and AI generally.» Papers based upon very small models, loose comparisons.» ALARM often used --- everything gets it to within 1 or 2 arcs.
Neil and Korb (1999) compared CaMML and BGe (Heckerman & Geiger's Bayesian metric over equivalence classes), using identical GA search over linear models:» On KL distance and topological distance from the true model,
CaMML and BGe performed nearly the same.» On test prediction accuracy on strict effect nodes (those with
no children), CaMML clearly outperformed BGe.
33
Extensions to original CaMML
Allow specification of prior on arc » O’Donnell, Korb, Nicholson» Useful for combining expert and automated
methods Learning local structure
» Logit models (Neill, Wallace, Korb)» Hybrid networks - CPT or decision trees
(O’Donnell, Allison, Korb, Hope) (Uses MCMC search)