8 - Bayesian Networks

Question 1

Bayesian Networks

Answer 1

Representation of conditional independence by a DAG

Question 2

What can be inferred from conditional independence by Bayesian networks?

Answer 2

Joint probability

Question 3

Monte Carlo Simulation

Answer 3

Approximation Technique

Question 4

Trade-off of Monte Carlo Simulation

Answer 4

Accuracy Inefficiency

Question 5

What are probabilities defining?

Answer 5

P. summarize the agent’s belief relative to the evidence.

Question 6

Why do we use Monte Carlo Simulation?

Answer 6

Exact inference computationally too expensive for large Bayesian Networks

Question 7

Disadvantage of Enumeration

Answer 7

Inefficiency due to repeated computation (Evaluation Tree)

Question 8

Independent Probability

Answer 8

No incoming arrows/probabilities (Start Node)

Question 9

Conditional Independence in logical terms

Answer 9

Cond. Ind.: P(X|Y;Z) = P(X|Z) [y dependent on z]

Question 10

Sample Space

Answer 10

Set of possible outcomes

Question 11

Event Space

Answer 11

Contains all possible combinations and outcomes

Question 12

Advantage of Likelihood Weighting

Answer 12

Avoids the inefficiency of rejection sampling by only generating events including the evidence e.

Question 13

Disadvantage of Rejection Sampling

Answer 13

Expensive for small P(e)

Question 14

What is the problem of exact inference in large Bayesian networks?

Answer 14

Computationally too Expensive

Question 15

Types of exact inference

Answer 15

Enumeration & Variable Elimination