DeepAI AI Chat
Log In Sign Up

Pretraining Representations for Data-Efficient Reinforcement Learning

by   Max Schwarzer, et al.
Montréal Institute of Learning Algorithms

Data efficiency is a key challenge for deep reinforcement learning. We address this problem by using unlabeled data to pretrain an encoder which is then finetuned on a small amount of task-specific data. To encourage learning representations which capture diverse aspects of the underlying MDP, we employ a combination of latent dynamics modelling and unsupervised goal-conditioned RL. When limited to 100k steps of interaction on Atari games (equivalent to two hours of human experience), our approach significantly surpasses prior work combining offline representation pretraining with task-specific finetuning, and compares favourably with other pretraining methods that require orders of magnitude more data. Our approach shows particular promise when combined with larger models as well as more diverse, task-aligned observational data – approaching human-level performance and data-efficiency on Atari in our best setting. We provide code associated with this work at


page 1

page 2

page 3

page 4


Pretraining Reinforcement Learning: Sharpening the Axe Before Cutting the Tree

Pretraining is a common technique in deep learning for increasing perfor...

Pretraining the Vision Transformer using self-supervised methods for vision based Deep Reinforcement Learning

The Vision Transformer architecture has shown to be competitive in the c...

A Compact Pretraining Approach for Neural Language Models

Domain adaptation for large neural language models (NLMs) is coupled wit...

Behavior Retrieval: Few-Shot Imitation Learning by Querying Unlabeled Datasets

Enabling robots to learn novel visuomotor skills in a data-efficient man...

APS: Active Pretraining with Successor Features

We introduce a new unsupervised pretraining objective for reinforcement ...

On Context Distribution Shift in Task Representation Learning for Offline Meta RL

Offline meta reinforcement learning (OMRL) aims to learn transferrable k...

Palm up: Playing in the Latent Manifold for Unsupervised Pretraining

Large and diverse datasets have been the cornerstones of many impressive...