Definition
Plain language
A diagram that captures how some facts depend on others, used to make probabilistic reasoning tractable.
As stated in the literature
A directed acyclic graph encoding conditional independence structure over a joint probability distribution, enabling efficient inference and motivating factored representations.
Also called: Bayesian networks
Why it matters: By making conditional independence explicit, Bayesian networks turn intractable probabilistic reasoning into something computers can actually solve.
For example, a Bayesian network for medical diagnosis might encode that a fever depends on infection, which depends on exposure, and use that structure to compute the chance of infection given an observed fever.
Heard on the show
“… local pieces, exploiting independence to make storage tractable — that's the move that made Bayesian networks tractable in the nineteen-eighties. …”Episode 042 — An Agentic Scientific Computing System That Actually Remembers What It Learns