Marginal Distribution
Definition
The marginal distribution of a subset of random variables is obtained by integrating (or summing) the joint distribution over the remaining variables.
- Continuous case:
- Discrete case:
Role in Inference
Computing marginals from a joint distribution is a fundamental operation in probabilistic graphical models. When the state space is large, exact marginalization becomes intractable, motivating approximate methods such as Belief Propagation (BP) and Variational Inference (VI).