Stable Hadamard Memory: Revitalizing Memory-Augmented Agents for Reinforcement Learning

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Le, Hung, Do, Kien, Nguyen, Dung, Gupta, Sunil, Venkatesh, Svetha
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912072130887680
author Le, Hung
Do, Kien
Nguyen, Dung
Gupta, Sunil
Venkatesh, Svetha
author_facet Le, Hung
Do, Kien
Nguyen, Dung
Gupta, Sunil
Venkatesh, Svetha
contents Effective decision-making in partially observable environments demands robust memory management. Despite their success in supervised learning, current deep-learning memory models struggle in reinforcement learning environments that are partially observable and long-term. They fail to efficiently capture relevant past information, adapt flexibly to changing observations, and maintain stable updates over long episodes. We theoretically analyze the limitations of existing memory models within a unified framework and introduce the Stable Hadamard Memory, a novel memory model for reinforcement learning agents. Our model dynamically adjusts memory by erasing no longer needed experiences and reinforcing crucial ones computationally efficiently. To this end, we leverage the Hadamard product for calibrating and updating memory, specifically designed to enhance memory capacity while mitigating numerical and learning challenges. Our approach significantly outperforms state-of-the-art memory-based methods on challenging partially observable benchmarks, such as meta-reinforcement learning, long-horizon credit assignment, and POPGym, demonstrating superior performance in handling long-term and evolving contexts.
format Preprint
id arxiv_https___arxiv_org_abs_2410_10132
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Stable Hadamard Memory: Revitalizing Memory-Augmented Agents for Reinforcement Learning
Le, Hung
Do, Kien
Nguyen, Dung
Gupta, Sunil
Venkatesh, Svetha
Machine Learning
Effective decision-making in partially observable environments demands robust memory management. Despite their success in supervised learning, current deep-learning memory models struggle in reinforcement learning environments that are partially observable and long-term. They fail to efficiently capture relevant past information, adapt flexibly to changing observations, and maintain stable updates over long episodes. We theoretically analyze the limitations of existing memory models within a unified framework and introduce the Stable Hadamard Memory, a novel memory model for reinforcement learning agents. Our model dynamically adjusts memory by erasing no longer needed experiences and reinforcing crucial ones computationally efficiently. To this end, we leverage the Hadamard product for calibrating and updating memory, specifically designed to enhance memory capacity while mitigating numerical and learning challenges. Our approach significantly outperforms state-of-the-art memory-based methods on challenging partially observable benchmarks, such as meta-reinforcement learning, long-horizon credit assignment, and POPGym, demonstrating superior performance in handling long-term and evolving contexts.
title Stable Hadamard Memory: Revitalizing Memory-Augmented Agents for Reinforcement Learning
topic Machine Learning
url https://arxiv.org/abs/2410.10132