Hierarchical Orchestra of Policies

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Cannon, Thomas P, Simsek, Özgür
Formato: Preprint
Publicado: 2024
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866910685655465984
author Cannon, Thomas P
Simsek, Özgür
author_facet Cannon, Thomas P
Simsek, Özgür
contents Continual reinforcement learning poses a major challenge due to the tendency of agents to experience catastrophic forgetting when learning sequential tasks. In this paper, we introduce a modularity-based approach, called Hierarchical Orchestra of Policies (HOP), designed to mitigate catastrophic forgetting in lifelong reinforcement learning. HOP dynamically forms a hierarchy of policies based on a similarity metric between the current observations and previously encountered observations in successful tasks. Unlike other state-of-the-art methods, HOP does not require task labelling, allowing for robust adaptation in environments where boundaries between tasks are ambiguous. Our experiments, conducted across multiple tasks in a procedurally generated suite of environments, demonstrate that HOP significantly outperforms baseline methods in retaining knowledge across tasks and performs comparably to state-of-the-art transfer methods that require task labelling. Moreover, HOP achieves this without compromising performance when tasks remain constant, highlighting its versatility.
format Preprint
id arxiv_https___arxiv_org_abs_2411_03008
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Hierarchical Orchestra of Policies
Cannon, Thomas P
Simsek, Özgür
Machine Learning
Artificial Intelligence
Continual reinforcement learning poses a major challenge due to the tendency of agents to experience catastrophic forgetting when learning sequential tasks. In this paper, we introduce a modularity-based approach, called Hierarchical Orchestra of Policies (HOP), designed to mitigate catastrophic forgetting in lifelong reinforcement learning. HOP dynamically forms a hierarchy of policies based on a similarity metric between the current observations and previously encountered observations in successful tasks. Unlike other state-of-the-art methods, HOP does not require task labelling, allowing for robust adaptation in environments where boundaries between tasks are ambiguous. Our experiments, conducted across multiple tasks in a procedurally generated suite of environments, demonstrate that HOP significantly outperforms baseline methods in retaining knowledge across tasks and performs comparably to state-of-the-art transfer methods that require task labelling. Moreover, HOP achieves this without compromising performance when tasks remain constant, highlighting its versatility.
title Hierarchical Orchestra of Policies
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2411.03008