Belief Network 0 nodes

Bayesian Networks

A Bayesian network is a probabilistic graphical model representing random variables and their conditional dependencies via a directed acyclic graph.

How to Use
  1. Set Evidence: Click a value label (left side) to observe that value
  2. Query: Click a percentage (right side) to see how it was calculated
  3. Watch posteriors update as you add evidence
Key Concepts
  • CPT: Conditional Probability Table (given values in each node)
  • Evidence: Observed values (green/red highlighting)
  • Posterior: P(Query | Evidence) - computed via inference

Things to Try

Alarm Network

A burglar or earthquake triggers an alarm, which may cause John and Mary to call.

  • Set John Calls = True. Watch P(Burglary) increase.
  • Also set Mary Calls = True. P(Burglary) increases more!
  • Now set Earthquake = True. P(Burglary) drops (explaining away).
Rain/Sprinkler

Classic "explaining away" example.

  • Set Wet Grass = True. Both P(Rain) and P(Sprinkler) increase.
  • Now set Rain = True. P(Sprinkler) decreases!

Inference Calculation

Click a percentage value in any node to see how that posterior probability is computed