Auto-Encoding Sequential Monte Carlo

Abstract

We build on auto-encoding sequential Monte Carlo (AESMC): a method for model and proposal learning based on maximizing the lower bound to the log marginal likelihood in a broad family of structured probabilistic models. Our approach relies on the efficiency of sequential Monte Carlo (SMC) for performing inference in structured probabilistic models and the flexibility of deep neural networks to model complex conditional probability distributions. We develop additional theoretical insights and introduce a new training procedure which improves both model and proposal learning. We demonstrate that our approach provides a fast, easy-to-implement and scalable means for simultaneous model learning and proposal adaptation in deep generative models.

Cited by in corpus (38)

• Variational Sequential Monte Carlo
• A Tutorial on Deep Latent Variable Models of Natural Language
• Tighter Variational Bounds are Not Necessarily Better
• Discriminative Particle Filter Reinforcement Learning for Complex Partial Observations
• Revisiting Reweighted Wake-Sleep for Models with Stochastic Control Flow
• Particle Filter Networks with Application to Visual Localization
• A Review of Learning with Deep Generative Models from Perspective of Graphical Modeling
• Markovian Score Climbing: Variational Inference with KL(p||q)
• Streaming Variational Monte Carlo
• Reweighted Expectation Maximization
• Adaptive Path-Integral Autoencoder: Representation Learning and Planning for Dynamical Systems
• Reparameterization Gradient for Non-differentiable Models
• Optimizing Sequential Medical Treatments with Auto-Encoding Heuristic Search in POMDPs
• Differentiable Particle Filtering without Modifying the Forward Pass
• Adaptive Monte Carlo augmented with normalizing flows
• Learning Proposals for Probabilistic Programs with Inference Combinators
• Particle Smoothing Variational Objectives
• An Easy to Interpret Diagnostic for Approximate Inference: Symmetric Divergence Over Simulations
• AIDE: An algorithm for measuring the accuracy of probabilistic inference algorithms
• Particle Filter Recurrent Neural Networks
• On importance-weighted autoencoders
• Annealed Flow Transport Monte Carlo
• The Thermodynamic Variational Objective
• Learning to Assimilate in Chaotic Dynamical Systems
• Filtering Variational Objectives
• Customizing Sequence Generation with Multi-Task Dynamical Systems
• Variational Dynamic Mixtures
• Optimized Auxiliary Particle Filters: adapting mixture proposals via convex optimization
• Variational Marginal Particle Filters
• Monte Carlo Filtering Objectives: A New Family of Variational Objectives to Learn Generative Model and Neural Adaptive Proposal for Time Series
• Energy-Inspired Models: Learning with Sampler-Induced Distributions
• Relational State-Space Model for Stochastic Multi-Object Systems
• Black-Box Autoregressive Density Estimation for State-Space Models
• Ensemble Kalman Variational Objectives: Nonlinear Latent Trajectory Inference with A Hybrid of Variational Inference and Ensemble Kalman Filter
• Online Variational Filtering and Parameter Learning
• Differentiable Annealed Importance Sampling and the Perils of Gradient Noise
• MCMC-Interactive Variational Inference
• Variational Combinatorial Sequential Monte Carlo Methods for Bayesian Phylogenetic Inference