Abstract:Building and maintaining state to learn policies and value functions is critical for deploying reinforcement learning (RL) agents in the real world. Recurrent neural networks (RNNs) have become a key point of interest for the state-building problem, and several large-scale reinforcement learning agents incorporate recurrent networks. While RNNs have become a mainstay in many RL applications, many key design choices and implementation details responsible for performance improvements are often not reported. In this work, we discuss one axis on which RNN architectures can be (and have been) modified for use in RL. Specifically, we look at how action information can be incorporated into the state update function of a recurrent cell. We discuss several choices in using action information and empirically evaluate the resulting architectures on a set of illustrative domains. Finally, we discuss future work in developing recurrent cells and discuss challenges specific to the RL setting.
| Comments: | Published in TMLR in 2023, https: // openreview. net/ forum? id= K6g4MbAC1r .Transactions on Machine Learning Research (2023) |
| Subjects: | Machine Learning (cs.LG) |
| Cite as: | arXiv:2605.16318 [cs.LG] |
| (or arXiv:2605.16318v1 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2605.16318 arXiv-issued DOI via DataCite |
From: Matthew Schlegel [view email]
[v1]
Mon, 4 May 2026 22:18:05 UTC (14,036 KB)
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。