Distributed Prioritized Experience Replay

Dan Horgan, John Quan, David Budden, Gabriel Barth-Maron, Matteo Hessel, Hado van Hasselt, David Silver

Abstract: We propose a distributed architecture for deep reinforcement learning at scale, that enables agents to learn effectively from orders of magnitude more data than previously possible. The algorithm decouples acting from learning: the actors interact with their own instances of the environment by selecting actions according to a shared neural network, and accumulate the resulting experience in a shared experience replay memory; the learner replays samples of experience and updates the neural network. The architecture relies on prioritized experience replay to focus only on the most significant data generated by the actors. Our architecture substantially improves the state of the art on the Arcade Learning Environment, achieving better final performance in a fraction of the wall-clock training time.

TL;DR: A distributed architecture for deep reinforcement learning at scale, using parallel data-generation to improve the state of the art on the Arcade Learning Environment benchmark in a fraction of the wall-clock training time of previous approaches.

Keywords: deep learning, reinforcement learning, distributed systems

Code: [ 15 community implementations](https://paperswithcode.com/paper/?openreview=H1Dy---0Z)

Data: [Arcade Learning Environment](https://paperswithcode.com/dataset/arcade-learning-environment)

Community Implementations: [ 13 code implementations](https://www.catalyzex.com/paper/arxiv:1803.00933/code)