Saved in:
Bibliographic Details
Main Authors: Hosu, Ionel-Alexandru, Rebedea, Traian, Pascanu, Razvan
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2506.13690
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912432925966336
author Hosu, Ionel-Alexandru
Rebedea, Traian
Pascanu, Razvan
author_facet Hosu, Ionel-Alexandru
Rebedea, Traian
Pascanu, Razvan
contents One proposed mechanism to improve exploration in reinforcement learning is through the use of macro-actions. Paradoxically though, in many scenarios the naive addition of macro-actions does not lead to better exploration, but rather the opposite. It has been argued that this was caused by adding non-useful macros and multiple works have focused on mechanisms to discover effectively environment-specific useful macros. In this work, we take a slightly different perspective. We argue that the difficulty stems from the trade-offs between reducing the average number of decisions per episode versus increasing the size of the action space. Namely, one typically treats each potential macro-action as independent and atomic, hence strictly increasing the search space and making typical exploration strategies inefficient. To address this problem we propose a novel regularization term that exploits the relationship between actions and macro-actions to improve the credit assignment mechanism by reducing the effective dimension of the action space and, therefore, improving exploration. The term relies on a similarity matrix that is meta-learned jointly with learning the desired policy. We empirically validate our strategy looking at macro-actions in Atari games, and the StreetFighter II environment. Our results show significant improvements over the Rainbow-DQN baseline in all environments. Additionally, we show that the macro-action similarity is transferable to related environments. We believe this work is a small but important step towards understanding how the similarity-imposed geometry on the action space can be exploited to improve credit assignment and exploration, therefore making learning more effective.
format Preprint
id arxiv_https___arxiv_org_abs_2506_13690
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Meta-learning how to Share Credit among Macro-Actions
Hosu, Ionel-Alexandru
Rebedea, Traian
Pascanu, Razvan
Machine Learning
Artificial Intelligence
One proposed mechanism to improve exploration in reinforcement learning is through the use of macro-actions. Paradoxically though, in many scenarios the naive addition of macro-actions does not lead to better exploration, but rather the opposite. It has been argued that this was caused by adding non-useful macros and multiple works have focused on mechanisms to discover effectively environment-specific useful macros. In this work, we take a slightly different perspective. We argue that the difficulty stems from the trade-offs between reducing the average number of decisions per episode versus increasing the size of the action space. Namely, one typically treats each potential macro-action as independent and atomic, hence strictly increasing the search space and making typical exploration strategies inefficient. To address this problem we propose a novel regularization term that exploits the relationship between actions and macro-actions to improve the credit assignment mechanism by reducing the effective dimension of the action space and, therefore, improving exploration. The term relies on a similarity matrix that is meta-learned jointly with learning the desired policy. We empirically validate our strategy looking at macro-actions in Atari games, and the StreetFighter II environment. Our results show significant improvements over the Rainbow-DQN baseline in all environments. Additionally, we show that the macro-action similarity is transferable to related environments. We believe this work is a small but important step towards understanding how the similarity-imposed geometry on the action space can be exploited to improve credit assignment and exploration, therefore making learning more effective.
title Meta-learning how to Share Credit among Macro-Actions
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2506.13690