Learning State Representations from Random Deep Action-conditional Predictions

Abstract

Our main contribution in this work is an empirical finding that random General Value Functions (GVFs), i.e., deep action-conditional predictions -- random both in what feature of observations they predict as well as in the sequence of actions the predictions are conditioned upon -- form good auxiliary tasks for reinforcement learning (RL) problems. In particular, we show that random deep action-conditional predictions when used as auxiliary tasks yield state representations that produce control performance competitive with state-of-the-art hand-crafted auxiliary tasks like value prediction, pixel control, and CURL in both Atari and DeepMind Lab tasks. In another set of experiments we stop the gradients from the RL part of the network to the state representation learning part of the network and show, perhaps surprisingly, that the auxiliary tasks alone are sufficient to learn state representations good enough to outperform an end-to-end trained actor-critic baseline. We opensourced our code at https://github.com/Hwhitetooth/random_gvfs.

arXiv commentNeurIPS 2021

References in corpus (9)

• Empirical Evaluation of Gated Recurrent Neural Networks on Sequence Modeling
• Reinforcement Learning with Unsupervised Auxiliary Tasks
• Hyperbolic Discounting and Learning over Multiple Horizons
• Data-Efficient Reinforcement Learning with Self-Predictive Representations
• Bootstrap Latent-Predictive Representations for Multitask Reinforcement Learning
• Discovery of Useful Questions as Auxiliary Tasks
• Deep Reinforcement and InfoMax Learning
• Mastering Atari with Discrete World Models
• General Value Function Networks