Automatic Rao-Blackwellization for Sequential Monte Carlo with Belief Propagation
This work addresses the problem of improving inference accuracy for complex models in machine learning and statistics, though it appears incremental as it builds on existing SMC and belief propagation techniques.
The paper tackles the intractability of exact Bayesian inference in state-space models by proposing a mixed inference algorithm that combines belief propagation for closed-form solutions with sampling-based SMC methods when exact computations fail, achieving exact results for Gaussian tree models.
Exact Bayesian inference on state-space models~(SSM) is in general untractable, and unfortunately, basic Sequential Monte Carlo~(SMC) methods do not yield correct approximations for complex models. In this paper, we propose a mixed inference algorithm that computes closed-form solutions using belief propagation as much as possible, and falls back to sampling-based SMC methods when exact computations fail. This algorithm thus implements automatic Rao-Blackwellization and is even exact for Gaussian tree models.