Variational Message Passing with Structured Inference Networks
This work provides a more interpretable and efficient inference method for deep structured models, though it appears incremental relative to existing variational auto-encoder approaches.
The paper tackles the challenge of combining deep models with probabilistic graphical models by proposing a variational message-passing algorithm that incorporates graphical model structure into inference networks, enabling efficient amortized and natural-gradient inference.
Recent efforts on combining deep models with probabilistic graphical models are promising in providing flexible models that are also easy to interpret. We propose a variational message-passing algorithm for variational inference in such models. We make three contributions. First, we propose structured inference networks that incorporate the structure of the graphical model in the inference network of variational auto-encoders (VAE). Second, we establish conditions under which such inference networks enable fast amortized inference similar to VAE. Finally, we derive a variational message passing algorithm to perform efficient natural-gradient inference while retaining the efficiency of the amortized inference. By simultaneously enabling structured, amortized, and natural-gradient inference for deep structured models, our method simplifies and generalizes existing methods.