Enregistré dans:
Détails bibliographiques
Auteurs principaux: Tolguenec, Paul-Antoine Le, Besse, Yann, Teichteil-Koenigsbuch, Florent, Wilson, Dennis G., Rachelson, Emmanuel
Format: Preprint
Publié: 2024
Sujets:
Accès en ligne:https://arxiv.org/abs/2411.14085
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866913582597275648
author Tolguenec, Paul-Antoine Le
Besse, Yann
Teichteil-Koenigsbuch, Florent
Wilson, Dennis G.
Rachelson, Emmanuel
author_facet Tolguenec, Paul-Antoine Le
Besse, Yann
Teichteil-Koenigsbuch, Florent
Wilson, Dennis G.
Rachelson, Emmanuel
contents The ability to explore efficiently and effectively is a central challenge of reinforcement learning. In this work, we consider exploration through the lens of information theory. Specifically, we cast exploration as a problem of maximizing the Shannon entropy of the state occupation measure. This is done by maximizing a sequence of divergences between distributions representing an agent's past behavior and its current behavior. Intuitively, this encourages the agent to explore new behaviors that are distinct from past behaviors. Hence, we call our method RAMP, for ``$\textbf{R}$unning $\textbf{A}$way fro$\textbf{m}$ the $\textbf{P}$ast.'' A fundamental question of this method is the quantification of the distribution change over time. We consider both the Kullback-Leibler divergence and the Wasserstein distance to quantify divergence between successive state occupation measures, and explain why the former might lead to undesirable exploratory behaviors in some tasks. We demonstrate that by encouraging the agent to explore by actively distancing itself from past experiences, it can effectively explore mazes and a wide range of behaviors on robotic manipulation and locomotion tasks.
format Preprint
id arxiv_https___arxiv_org_abs_2411_14085
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Exploration by Running Away from the Past
Tolguenec, Paul-Antoine Le
Besse, Yann
Teichteil-Koenigsbuch, Florent
Wilson, Dennis G.
Rachelson, Emmanuel
Machine Learning
The ability to explore efficiently and effectively is a central challenge of reinforcement learning. In this work, we consider exploration through the lens of information theory. Specifically, we cast exploration as a problem of maximizing the Shannon entropy of the state occupation measure. This is done by maximizing a sequence of divergences between distributions representing an agent's past behavior and its current behavior. Intuitively, this encourages the agent to explore new behaviors that are distinct from past behaviors. Hence, we call our method RAMP, for ``$\textbf{R}$unning $\textbf{A}$way fro$\textbf{m}$ the $\textbf{P}$ast.'' A fundamental question of this method is the quantification of the distribution change over time. We consider both the Kullback-Leibler divergence and the Wasserstein distance to quantify divergence between successive state occupation measures, and explain why the former might lead to undesirable exploratory behaviors in some tasks. We demonstrate that by encouraging the agent to explore by actively distancing itself from past experiences, it can effectively explore mazes and a wide range of behaviors on robotic manipulation and locomotion tasks.
title Exploration by Running Away from the Past
topic Machine Learning
url https://arxiv.org/abs/2411.14085