Computational Perception Group

home | members | publications | directions | research | presentations | teaching | vacancies | personal

Neural Sequential Monte Carlo

Shixiang (Shane) Gu, Zoubin Ghahramani and Richard E. Turner

Published in: Advances in Neural Information Processing Systems

Sequential Monte Carlo (SMC), or particle filtering, is a popular class of meth- ods for sampling from an intractable target distribution using a sequence of sim- pler intermediate distributions. Like other importance sampling-based methods, performance is critically dependent on the proposal distribution: a bad proposal can lead to arbitrarily inaccurate estimates of the target distribution. This paper presents a new method for automatically adapting the proposal using an approx- imation of the Kullback-Leibler divergence between the true posterior and the proposal distribution. The method is very flexible, applicable to any parameter- ized proposal distribution and it supports online and batch variants. We use the new framework to adapt powerful proposal distributions with rich parameteriza- tions based upon neural networks leading to Neural Adaptive Sequential Monte Carlo (NASMC). Experiments indicate that NASMC significantly improves infer- ence in a non-linear state space model outperforming adaptive proposal methods including the Extended Kalman and Unscented Particle Filters. Experiments also indicate that improved inference translates into improved parameter learning when NASMC is used as a subroutine of Particle Marginal Metropolis Hastings. Finally we show that NASMC is able to train a latent variable recurrent neural network (LV-RNN) achieving results that compete with the state-of-the-art for polymor- phic music modelling. NASMC can be seen as bridging the gap between adaptive SMC methods and the recent work in scalable, black-box variational inference.

bibtex, pdf, supplementary material

This site is powered by FoswikiCopyright © by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding Foswiki? Send feedback