English

Fast deep reinforcement learning using online adjustments from the past

Machine Learning 2018-10-19 v1 Artificial Intelligence

Abstract

We propose Ephemeral Value Adjusments (EVA): a means of allowing deep reinforcement learning agents to rapidly adapt to experience in their replay buffer. EVA shifts the value predicted by a neural network with an estimate of the value function found by planning over experience tuples from the replay buffer near the current state. EVA combines a number of recent ideas around combining episodic memory-like structures into reinforcement learning agents: slot-based storage, content-based retrieval, and memory-based planning. We show that EVAis performant on a demonstration task and Atari games.

Keywords

Cite

@article{arxiv.1810.08163,
  title  = {Fast deep reinforcement learning using online adjustments from the past},
  author = {Steven Hansen and Pablo Sprechmann and Alexander Pritzel and André Barreto and Charles Blundell},
  journal= {arXiv preprint arXiv:1810.08163},
  year   = {2018}
}

Comments

Accepted at NIPS 2018

R2 v1 2026-06-23T04:44:52.135Z