next up previous
Next: Inference. Up: Marginalization, inference and sampling Previous: Marginalization, inference and sampling

Marginalization.

The marginal distribution of a subset $A \subset V$ is given as follows:

\begin{displaymath}
Q_A(x_A)\;=\;\sum_{k=1}^m \lambda_k T^k_A( x_A).
\end{displaymath}

Hence, the marginal of $Q$ is a mixture of the marginals of the component trees.

Journal of Machine Learning Research 2000-10-19