Variational Marginal Particle Filters
arXiv:2109.15134
Abstract
Variational inference for state space models (SSMs) is known to be hard in general. Recent works focus on deriving variational objectives for SSMs from unbiased sequential Monte Carlo estimators. We reveal that the marginal particle filter is obtained from sequential Monte Carlo by applying Rao-Blackwellization operations, which sacrifices the trajectory information for reduced variance and differentiability. We propose the variational marginal particle filter (VMPF), which is a differentiable and reparameterizable variational filtering objective for SSMs based on an unbiased estimator. We find that VMPF with biased gradients gives tighter bounds than previous objectives, and the unbiased reparameterization gradients are sometimes beneficial.
Accepted to AISTATS 2022
References in corpus (15)
- Categorical Reparameterization with Gumbel-Softmax
- The Concrete Distribution: A Continuous Relaxation of Discrete Random Variables
- Markov Chain Monte Carlo and Variational Inference: Bridging the Gap
- Implicit Reparameterization Gradients
- Auto-Encoding Sequential Monte Carlo
- Variational Boosting: Iteratively Refining Posterior Approximations
- Stochastic Backpropagation through Mixture Density Distributions
- Discriminative Particle Filter Reinforcement Learning for Complex Partial Observations
- Towards Differentiable Resampling
- Differentiable Particle Filtering via Entropy-Regularized Optimal Transport
- Tensor Monte Carlo: particle methods for the GPU era
- Learning Proposals for Probabilistic Programs with Inference Combinators
- Particle Smoothing Variational Objectives
- Divide and Couple: Using Monte Carlo Variational Objectives for Posterior Approximation
- Copula-like Variational Inference