Conditional Probability Network Instance

A conditional probability network instance, [math]\displaystyle{ P(X|Y_1,...,Y_n) }[/math], is a probabilistic network whose edges represent conditional probabilities such that the conditional probability of a node given all other nodes are equivalent to the conditional probability with all of its neighboring nodes.

• AKA: Conditional Probabilistic Graphical Model.
• Context:
  • It can be a compact representation of joint probabilities.
  • It can be used for Statistical Inference (to infer the Joint Probability Value for an Unobserved Random Variable given an Observed Random Variable).
  • It can be used to represent Domain Knowledge.
  • It can conform to a Conditional Probabilistic Graphical Metamodel.
  • It can range from being an Undirected Conditional Probability Network to being a Directed Conditional Probability Network.
  • It can (typically) be treated as a Discriminative Model.
  • …
• Example(s):
  • a Bayesian Network.
  • an Undirected Conditional Probability Network, such as a Conditional Random Field Network Instance.
  • …
• Counter-Example(s):
  • a Markov Logic Network.
• See: Hidden Markov Network.

---