TY - GEN

T1 - AutoRL Hyperparameter Landscapes

AU - Mohan, Aditya

AU - Benjamins, Carolin

AU - Wienecke, Konrad

AU - Dockhorn, Alexander

AU - Lindauer, Marius

N1 - Publisher Copyright: ©2023 the authors.

PY - 2023/11/12

Y1 - 2023/11/12

N2 - Although Reinforcement Learning (RL) has shown to be capable of producing impressive results, its use is limited by the impact of its hyperparameters on performance. This often makes it difficult to achieve good results in practice. Automated RL (AutoRL) addresses this difficulty, yet little is known about the dynamics of the hyperparameter landscapes that hyperparameter optimization (HPO) methods traverse in search of optimal configurations. In view of existing AutoRL approaches dynamically adjusting hyperparameter configurations, we propose an approach to build and analyze these hyperparameter landscapes not just for one point in time but at multiple points in time throughout training. Addressing an important open question on the legitimacy of such dynamic AutoRL approaches, we provide thorough empirical evidence that the hyperparameter landscapes strongly vary over time across representative algorithms from RL literature (DQN and SAC) in different kinds of environments (Cartpole and Hopper). This supports the theory that hyperparameters should be dynamically adjusted during training and shows the potential for more insights on AutoRL problems that can be gained through landscape analyses.

AB - Although Reinforcement Learning (RL) has shown to be capable of producing impressive results, its use is limited by the impact of its hyperparameters on performance. This often makes it difficult to achieve good results in practice. Automated RL (AutoRL) addresses this difficulty, yet little is known about the dynamics of the hyperparameter landscapes that hyperparameter optimization (HPO) methods traverse in search of optimal configurations. In view of existing AutoRL approaches dynamically adjusting hyperparameter configurations, we propose an approach to build and analyze these hyperparameter landscapes not just for one point in time but at multiple points in time throughout training. Addressing an important open question on the legitimacy of such dynamic AutoRL approaches, we provide thorough empirical evidence that the hyperparameter landscapes strongly vary over time across representative algorithms from RL literature (DQN and SAC) in different kinds of environments (Cartpole and Hopper). This supports the theory that hyperparameters should be dynamically adjusted during training and shows the potential for more insights on AutoRL problems that can be gained through landscape analyses.

KW - Reinforcement learning

KW - AutoML

KW - Hyperparameter optimization

UR - http://www.scopus.com/inward/record.url?scp=85184347459&partnerID=8YFLogxK

U2 - 10.48550/arXiv.2304.02396

DO - 10.48550/arXiv.2304.02396

M3 - Conference contribution

T3 - Proceedings of Machine Learning Research

BT - Conference proceeding

T2 - 2nd International Conference on Automated Machine Learning, AutoML 2023

Y2 - 12 November 2023 through 15 November 2023

ER -