Disentangling Transfer in Continual Reinforcement Learning
This work addresses the problem of improving transfer in continual learning for reinforcement learning practitioners, but it is incremental as it builds on existing methods and benchmarks.
The study tackled the challenge of knowledge transfer in continual reinforcement learning by systematically analyzing components of the SAC algorithm, resulting in ClonEx-SAC achieving an 87% final success rate compared to 80% for the best existing method and improving transfer from 0.18 to 0.54.
The ability of continual learning systems to transfer knowledge from previously seen tasks in order to maximize performance on new tasks is a significant challenge for the field, limiting the applicability of continual learning solutions to realistic scenarios. Consequently, this study aims to broaden our understanding of transfer and its driving forces in the specific case of continual reinforcement learning. We adopt SAC as the underlying RL algorithm and Continual World as a suite of continuous control tasks. We systematically study how different components of SAC (the actor and the critic, exploration, and data) affect transfer efficacy, and we provide recommendations regarding various modeling options. The best set of choices, dubbed ClonEx-SAC, is evaluated on the recent Continual World benchmark. ClonEx-SAC achieves 87% final success rate compared to 80% of PackNet, the best method in the benchmark. Moreover, the transfer grows from 0.18 to 0.54 according to the metric provided by Continual World.