Adaptive Path-Integral Autoencoder: Representation Learning and Planning for Dynamical Systems

TL;DR

This paper introduces an adaptive path-integral autoencoder that learns low-dimensional latent dynamical systems from high-dimensional sequential data, enabling effective prediction and planning through a novel control-inference duality approach.

Contribution

It proposes a new variational inference framework combining amortized inference and path integral control to learn and utilize latent dynamical models for sequential data.

Findings

Tighter lower bounds in variational inference for sequential data.

Effective prediction and planning in learned low-dimensional latent spaces.

Demonstrated success on high-dimensional sequential datasets.

Abstract

We present a representation learning algorithm that learns a low-dimensional latent dynamical system from high-dimensional \textit{sequential} raw data, e.g., video. The framework builds upon recent advances in amortized inference methods that use both an inference network and a refinement procedure to output samples from a variational distribution given an observation sequence, and takes advantage of the duality between control and inference to approximately solve the intractable inference problem using the path integral control approach. The learned dynamical model can be used to predict and plan the future states; we also present the efficient planning method that exploits the learned low-dimensional latent dynamics. Numerical experiments show that the proposed path-integral control based variational inference method leads to tighter lower bounds in statistical model learning of sequential data. The supplementary video: https://youtu.be/xCp35crUoLQ