Bayesian Belief Networks

Question 1

Bayesian methods offers an approach to analysis, whereby ______ is incorporated into analysis through ______.

Answer 1

Bayesian methods offers an approach to analysis, whereby EXPERT JUDGEMENT is incorporated into analysis through Subjective Probability Distributions (also known as prior distributions, since they are obtained prior to any data being observed).

Question 2

Bayes Theorem is concerned with ______.

Answer 2

Bayes Theorem is concerned with conditional probabilities.

Question 3

Bayesian Belief Networks (BBN) form a framework for ______ → ______.
Therefore this method combines ______ and ______ data.

Answer 3

Bayesian Belief Networks (BBN) form a framework for decision support → reason.
Therefore this method combines subjective and objective data.

Question 4

A BBN _____ represents the relationship between variables, by capturing the _____ in dependencies between variables.

Answer 4

A BBN graphically represents the relationship between variables, by capturing the uncertainty in dependencies between variables.

Question 5

Mathematically, the BBN structure only reveals the _____ between variables.

Answer 5

Mathematically, the BBN structure only reveals the conditional independence between variables.

Question 6

Practically, the BBN captures the _____ between variables.

It uses _____ probabilities to represent the degree of belief in these relationships.

Answer 6

Practically, the BBN captures the perceived relationships between variables.

It uses conditional probabilities to represent the degree of belief in these relationships.

Question 7

What are the (brief) 7 steps to building a BBN?

Answer 7

1. Identify the variables
2. Network (directed acyclic graph)
3. Diagrammatic representation of casual relationships
4. Conditional Probability tables
5. Measurements of influences between variables
6. Enter evidence
7. Propagate evidence through network

Question 8

Give some strengths of BBNs

Answer 8

• Easy to understand
• Dependencies between variables are made explicit
• Separate sources of evidence can be combined into one framework
• Uncertainty in reasoning is taken into account
• Increasingly powerful software makes probabilistic reasoning relatively easy

Question 9

Give some weaknesses of BBNs

Answer 9

• Causation – assumes conditional (statistical) independence between variables
• Computational limitations with respect to continuous variables
• If the BBN model is grounded in subject-matter knowledge (in whole or part), then we need a method which can collect subjective judgement