AI · Theory

Bayesian Belief Network

Theory

46/215

Bayesian Belief Network

A Bayesian network is a directed graph: nodes are variables, edges are direct influence, each node has a CPT. Joint probability factors into local tables. Diagnosis and spam stacks use this idea. Independent variables have no edge.

Don’t draw a fully connected mess. A tiny Rain → Sprinkler / Rain → WetGrass sketch is enough.

Output — idea / do / check for Bayesian Belief Network. Fill those three in the viva.

Diagram

prior P(H)
       × P(E|H)
       / P(E)
       ▼
  posterior P(H|E)

Exam tip

Rain–sprinkler–grass sketch.

Example

# Bayesian Belief Network
print("idea :", "Bayesian Belief Network")
print("do   : one tiny example on paper")
print("check: one failure case")

Output: idea / do / check for Bayesian Belief Network. Fill those three in the viva.

Short notes

DefDAG + CPTs.
RuleLocal tables, not one giant joint.

AI · Theory

Bayesian Belief Network

← All stacks

Theory

46/215

Bayesian Belief Network

Don’t draw a fully connected mess. A tiny Rain → Sprinkler / Rain → WetGrass sketch is enough.

Output — idea / do / check for Bayesian Belief Network. Fill those three in the viva.

Diagram

prior P(H)
       × P(E|H)
       / P(E)
       ▼
  posterior P(H|E)

Exam tip

Rain–sprinkler–grass sketch.

Example

# Bayesian Belief Network
print("idea :", "Bayesian Belief Network")
print("do   : one tiny example on paper")
print("check: one failure case")

Output: idea / do / check for Bayesian Belief Network. Fill those three in the viva.

Short notes

DefDAG + CPTs.
RuleLocal tables, not one giant joint.

Bayesian Belief Network

Example

Short notes

Questions

Bayesian Belief Network

Example

Short notes

Questions