Measuring and Mitigating Interference in Reinforcement Learning

Vincent Liu, Han Wang, Ruo Yu Tao, Khurram Javed, Adam White, Martha White
Proceedings of The 2nd Conference on Lifelong Learning Agents, PMLR 232:781-795, 2023.

Abstract

Catastrophic interference is common in many network-based learning systems, and many proposals exist for mitigating it. Before overcoming interference we must understand it better. In this work, we provide a definition and novel measure of interference for value-based reinforcement learning methods such as Fitted Q-Iteration and DQN. We systematically evaluate our measure of interference, showing that it correlates with instability in control performance, across a variety of network architectures. Our new interference measure allows us to ask novel scientific questions about commonly used deep learning architectures and study learning algorithms which mitigate interference. Lastly, we outline a class of algorithms which we call online-aware that are designed to mitigate interference, and show they do reduce interference according to our measure and that they improve stability and performance in several classic control environments.

Cite this Paper

BibTeX
@InProceedings{pmlr-v232-liu23a, title = {Measuring and Mitigating Interference in Reinforcement Learning}, author = {Liu, Vincent and Wang, Han and Tao, Ruo Yu and Javed, Khurram and White, Adam and White, Martha}, booktitle = {Proceedings of The 2nd Conference on Lifelong Learning Agents}, pages = {781--795}, year = {2023}, editor = {Chandar, Sarath and Pascanu, Razvan and Sedghi, Hanie and Precup, Doina}, volume = {232}, series = {Proceedings of Machine Learning Research}, month = {22--25 Aug}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v232/liu23a/liu23a.pdf}, url = {https://proceedings.mlr.press/v232/liu23a.html}, abstract = {Catastrophic interference is common in many network-based learning systems, and many proposals exist for mitigating it. Before overcoming interference we must understand it better. In this work, we provide a definition and novel measure of interference for value-based reinforcement learning methods such as Fitted Q-Iteration and DQN. We systematically evaluate our measure of interference, showing that it correlates with instability in control performance, across a variety of network architectures. Our new interference measure allows us to ask novel scientific questions about commonly used deep learning architectures and study learning algorithms which mitigate interference. Lastly, we outline a class of algorithms which we call online-aware that are designed to mitigate interference, and show they do reduce interference according to our measure and that they improve stability and performance in several classic control environments.} }
Endnote
%0 Conference Paper %T Measuring and Mitigating Interference in Reinforcement Learning %A Vincent Liu %A Han Wang %A Ruo Yu Tao %A Khurram Javed %A Adam White %A Martha White %B Proceedings of The 2nd Conference on Lifelong Learning Agents %C Proceedings of Machine Learning Research %D 2023 %E Sarath Chandar %E Razvan Pascanu %E Hanie Sedghi %E Doina Precup %F pmlr-v232-liu23a %I PMLR %P 781--795 %U https://proceedings.mlr.press/v232/liu23a.html %V 232 %X Catastrophic interference is common in many network-based learning systems, and many proposals exist for mitigating it. Before overcoming interference we must understand it better. In this work, we provide a definition and novel measure of interference for value-based reinforcement learning methods such as Fitted Q-Iteration and DQN. We systematically evaluate our measure of interference, showing that it correlates with instability in control performance, across a variety of network architectures. Our new interference measure allows us to ask novel scientific questions about commonly used deep learning architectures and study learning algorithms which mitigate interference. Lastly, we outline a class of algorithms which we call online-aware that are designed to mitigate interference, and show they do reduce interference according to our measure and that they improve stability and performance in several classic control environments.
APA
Liu, V., Wang, H., Tao, R.Y., Javed, K., White, A. & White, M.. (2023). Measuring and Mitigating Interference in Reinforcement Learning. Proceedings of The 2nd Conference on Lifelong Learning Agents, in Proceedings of Machine Learning Research 232:781-795 Available from https://proceedings.mlr.press/v232/liu23a.html.

Related Material