Don't Do What Doesn't Matter: Intrinsic Motivation with Action Usefulness - Inria - Institut national de recherche en sciences et technologies du numérique Access content directly
Conference Papers Year : 2021

Don't Do What Doesn't Matter: Intrinsic Motivation with Action Usefulness

Abstract

Sparse rewards are double-edged training signals in reinforcement learning: easy to design but hard to optimize. Intrinsic motivation guidances have thus been developed toward alleviating the resulting exploration problem. They usually incentivize agents to look for new states through novelty signals. Yet, such methods encourage exhaustive exploration of the state space rather than focusing on the environment's salient interaction opportunities. We propose a new exploration method, called Don't Do What Doesn't Matter (DoWhaM), shifting the emphasis from state novelty to state with relevant actions. While most actions consistently change the state when used, e.g. moving the agent, some actions are only effective in specific states, e.g., opening a door, grabbing an object. DoWhaM detects and rewards actions that seldom affect the environment. We evaluate DoWhaM on the procedurallygenerated environment MiniGrid, against state-ofthe-art methods. Experiments consistently show that DoWhaM greatly reduces sample complexity, installing the new state-of-the-art in MiniGrid.
Fichier principal
Vignette du fichier
Rare_Actions_Matter_IJCAI.pdf (2.97 Mo) Télécharger le fichier
Origin : Files produced by the author(s)

Dates and versions

hal-03259315 , version 1 (14-06-2021)

Identifiers

  • HAL Id : hal-03259315 , version 1

Cite

Mathieu Seurin, Florian Strub, Philippe Preux, Olivier Pietquin. Don't Do What Doesn't Matter: Intrinsic Motivation with Action Usefulness. Internationnal Joint Conference on Artificial Intelligence (IJCAI), Aug 2021, Montreal, Canada. pp.2950--2956. ⟨hal-03259315⟩
79 View
99 Download

Share

Gmail Facebook X LinkedIn More