Representation Matters: Offline Pretraining for Sequential Decision Making

TL;DR

This paper explores how offline unsupervised pretraining of state representations can significantly enhance the performance of various downstream sequential decision-making tasks, including RL, imitation learning, and offline policy optimization.

Contribution

It introduces the use of offline unsupervised learning objectives for state representation pretraining to improve downstream decision-making tasks, a novel approach in offline RL.

Findings

Pretraining with unsupervised objectives boosts downstream policy performance.

Reward prediction and representation type are critical components.

Pretraining benefits are consistent across different offline datasets.

Abstract

The recent success of supervised learning methods on ever larger offline datasets has spurred interest in the reinforcement learning (RL) field to investigate whether the same paradigms can be translated to RL algorithms. This research area, known as offline RL, has largely focused on offline policy optimization, aiming to find a return-maximizing policy exclusively from offline data. In this paper, we consider a slightly different approach to incorporating offline data into sequential decision-making. We aim to answer the question, what unsupervised objectives applied to offline datasets are able to learn state representations which elevate performance on downstream tasks, whether those downstream tasks be online RL, imitation learning from expert demonstrations, or even offline policy optimization based on the same offline dataset? Through a variety of experiments utilizing standard offline RL datasets, we find that the use of pretraining with unsupervised learning objectives can dramatically improve the performance of policy learning algorithms that otherwise yield mediocre performance on their own. Extensive ablations further provide insights into what components of these unsupervised objectives -- e.g., reward prediction, continuous or discrete representations, pretraining or finetuning -- are most important and in which settings.