Skip to Main content Skip to Navigation
Conference papers

Adversarially Guided Actor-Critic

Yannis Flet-Berliac 1 Johan Ferret 2, 1 Olivier Pietquin 2 Philippe Preux 1 Matthieu Geist 2
1 Scool - Scool
Inria Lille - Nord Europe, CRIStAL - Centre de Recherche en Informatique, Signal et Automatique de Lille - UMR 9189
Abstract : Despite definite success in deep reinforcement learning problems, actor-critic algorithms are still confronted with sample inefficiency in complex environments, particularly in tasks where efficient exploration is a bottleneck. These methods consider a policy (the actor) and a value function (the critic) whose respective losses are built using different motivations and approaches. This paper introduces a third protagonist: the adversary. While the adversary mimics the actor by minimizing the KL-divergence between their respective action distributions, the actor, in addition to learning to solve the task, tries to differentiate itself from the adversary predictions. This novel objective stimulates the actor to follow strategies that could not have been correctly predicted from previous trajectories, making its behavior innovative in tasks where the reward is extremely rare. Our experimental analysis shows that the resulting Adversarially Guided Actor-Critic (AGAC) algorithm leads to more exhaustive exploration. Notably, AGAC outperforms current state-of-the-art methods on a set of various hard-exploration and procedurally-generated tasks.
Complete list of metadata

https://hal.inria.fr/hal-03167169
Contributor : Yannis Flet-Berliac <>
Submitted on : Thursday, March 11, 2021 - 6:55:57 PM
Last modification on : Monday, March 15, 2021 - 12:14:14 PM

File

AGAC.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : hal-03167169, version 1
  • ARXIV : 2102.04376

Collections

Citation

Yannis Flet-Berliac, Johan Ferret, Olivier Pietquin, Philippe Preux, Matthieu Geist. Adversarially Guided Actor-Critic. ICLR 2021 - International Conference on Learning Representations, May 2021, Vienna / Virtual, Austria. ⟨hal-03167169⟩

Share

Metrics

Record views

24

Files downloads

192