Tensor Variable Elimination for Plated Factor Graphs
This work addresses the need for compact representation and efficient inference in probabilistic models with repeated structures, such as in music modeling and sentiment analysis, though it is incremental by extending existing factor graph methods.
The authors tackled the problem of expressing repeated structure in factor graphs by introducing plated factor graphs and a tensor variable elimination algorithm, enabling exact inference in discrete latent variable models with repeated structure and integrating it into Pyro for validation on various graphical models.
A wide class of machine learning algorithms can be reduced to variable elimination on factor graphs. While factor graphs provide a unifying notation for these algorithms, they do not provide a compact way to express repeated structure when compared to plate diagrams for directed graphical models. To exploit efficient tensor algebra in graphs with plates of variables, we generalize undirected factor graphs to plated factor graphs and variable elimination to a tensor variable elimination algorithm that operates directly on plated factor graphs. Moreover, we generalize complexity bounds based on treewidth and characterize the class of plated factor graphs for which inference is tractable. As an application, we integrate tensor variable elimination into the Pyro probabilistic programming language to enable exact inference in discrete latent variable models with repeated structure. We validate our methods with experiments on both directed and undirected graphical models, including applications to polyphonic music modeling, animal movement modeling, and latent sentiment analysis.