Generative Memory for Lifelong Reinforcement Learning
This addresses the problem of maintaining performance across tasks for AI systems in lifelong learning scenarios, though it appears incremental as it builds on existing memory consolidation and replay techniques.
The paper tackles catastrophic forgetting in lifelong reinforcement learning by proposing a generative memory system that uses batch recall and pseudo-rehearsal to train multi-task agents, showing that task-agnostic latent space separation improves performance.
Our research is focused on understanding and applying biological memory transfers to new AI systems that can fundamentally improve their performance, throughout their fielded lifetime experience. We leverage current understanding of biological memory transfer to arrive at AI algorithms for memory consolidation and replay. In this paper, we propose the use of generative memory that can be recalled in batch samples to train a multi-task agent in a pseudo-rehearsal manner. We show results motivating the need for task-agnostic separation of latent space for the generative memory to address issues of catastrophic forgetting in lifelong learning.