Selective Experience Replay for Lifelong Learning
This addresses the problem of forgetting in lifelong learning for AI systems, but it is incremental as it builds on existing experience replay methods.
The paper tackles catastrophic forgetting in deep reinforcement learning when learning multiple tasks sequentially by proposing selective experience replay with four strategies, showing that distribution matching consistently prevents forgetting and performs best across domains.
Deep reinforcement learning has emerged as a powerful tool for a variety of learning tasks, however deep nets typically exhibit forgetting when learning multiple tasks in sequence. To mitigate forgetting, we propose an experience replay process that augments the standard FIFO buffer and selectively stores experiences in a long-term memory. We explore four strategies for selecting which experiences will be stored: favoring surprise, favoring reward, matching the global training distribution, and maximizing coverage of the state space. We show that distribution matching successfully prevents catastrophic forgetting, and is consistently the best approach on all domains tested. While distribution matching has better and more consistent performance, we identify one case in which coverage maximization is beneficial - when tasks that receive less trained are more important. Overall, our results show that selective experience replay, when suitable selection algorithms are employed, can prevent catastrophic forgetting.