Stop Regressing: Training Value Functions via Classification for Scalable Deep RL

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Farebrother, Jesse, Orbay, Jordi, Vuong, Quan, Taïga, Adrien Ali, Chebotar, Yevgen, Xiao, Ted, Irpan, Alex, Levine, Sergey, Castro, Pablo Samuel, Faust, Aleksandra, Kumar, Aviral, Agarwal, Rishabh
Formato: Preprint
Publicado: 2024
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866910355497680896
author Farebrother, Jesse
Orbay, Jordi
Vuong, Quan
Taïga, Adrien Ali
Chebotar, Yevgen
Xiao, Ted
Irpan, Alex
Levine, Sergey
Castro, Pablo Samuel
Faust, Aleksandra
Kumar, Aviral
Agarwal, Rishabh
author_facet Farebrother, Jesse
Orbay, Jordi
Vuong, Quan
Taïga, Adrien Ali
Chebotar, Yevgen
Xiao, Ted
Irpan, Alex
Levine, Sergey
Castro, Pablo Samuel
Faust, Aleksandra
Kumar, Aviral
Agarwal, Rishabh
contents Value functions are a central component of deep reinforcement learning (RL). These functions, parameterized by neural networks, are trained using a mean squared error regression objective to match bootstrapped target values. However, scaling value-based RL methods that use regression to large networks, such as high-capacity Transformers, has proven challenging. This difficulty is in stark contrast to supervised learning: by leveraging a cross-entropy classification loss, supervised methods have scaled reliably to massive networks. Observing this discrepancy, in this paper, we investigate whether the scalability of deep RL can also be improved simply by using classification in place of regression for training value functions. We demonstrate that value functions trained with categorical cross-entropy significantly improves performance and scalability in a variety of domains. These include: single-task RL on Atari 2600 games with SoftMoEs, multi-task RL on Atari with large-scale ResNets, robotic manipulation with Q-transformers, playing Chess without search, and a language-agent Wordle task with high-capacity Transformers, achieving state-of-the-art results on these domains. Through careful analysis, we show that the benefits of categorical cross-entropy primarily stem from its ability to mitigate issues inherent to value-based RL, such as noisy targets and non-stationarity. Overall, we argue that a simple shift to training value functions with categorical cross-entropy can yield substantial improvements in the scalability of deep RL at little-to-no cost.
format Preprint
id arxiv_https___arxiv_org_abs_2403_03950
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Stop Regressing: Training Value Functions via Classification for Scalable Deep RL
Farebrother, Jesse
Orbay, Jordi
Vuong, Quan
Taïga, Adrien Ali
Chebotar, Yevgen
Xiao, Ted
Irpan, Alex
Levine, Sergey
Castro, Pablo Samuel
Faust, Aleksandra
Kumar, Aviral
Agarwal, Rishabh
Machine Learning
Artificial Intelligence
Value functions are a central component of deep reinforcement learning (RL). These functions, parameterized by neural networks, are trained using a mean squared error regression objective to match bootstrapped target values. However, scaling value-based RL methods that use regression to large networks, such as high-capacity Transformers, has proven challenging. This difficulty is in stark contrast to supervised learning: by leveraging a cross-entropy classification loss, supervised methods have scaled reliably to massive networks. Observing this discrepancy, in this paper, we investigate whether the scalability of deep RL can also be improved simply by using classification in place of regression for training value functions. We demonstrate that value functions trained with categorical cross-entropy significantly improves performance and scalability in a variety of domains. These include: single-task RL on Atari 2600 games with SoftMoEs, multi-task RL on Atari with large-scale ResNets, robotic manipulation with Q-transformers, playing Chess without search, and a language-agent Wordle task with high-capacity Transformers, achieving state-of-the-art results on these domains. Through careful analysis, we show that the benefits of categorical cross-entropy primarily stem from its ability to mitigate issues inherent to value-based RL, such as noisy targets and non-stationarity. Overall, we argue that a simple shift to training value functions with categorical cross-entropy can yield substantial improvements in the scalability of deep RL at little-to-no cost.
title Stop Regressing: Training Value Functions via Classification for Scalable Deep RL
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2403.03950