Lifetime policy reuse and the importance of task capacity
This addresses the problem of catastrophic forgetting and policy proliferation in lifelong reinforcement learning for AI systems, though it appears incremental as it builds on existing policy reuse techniques.
The paper tackles lifelong reinforcement learning by introducing Lifetime Policy Reuse, a model-agnostic algorithm that optimizes a fixed number of policies to avoid generating many, and demonstrates its importance on domains with up to 125 tasks.
A long-standing challenge in artificial intelligence is lifelong reinforcement learning, where learners are given many tasks in sequence and must transfer knowledge between tasks while avoiding catastrophic forgetting. Policy reuse and other multi-policy reinforcement learning techniques can learn multiple tasks but may generate many policies. This paper presents two novel contributions, namely 1) Lifetime Policy Reuse, a model-agnostic policy reuse algorithm that avoids generating many policies by optimising a fixed number of near-optimal policies through a combination of policy optimisation and adaptive policy selection; and 2) the task capacity, a measure for the maximal number of tasks that a policy can accurately solve. Comparing two state-of-the-art base-learners, the results demonstrate the importance of Lifetime Policy Reuse and task capacity based pre-selection on an 18-task partially observable Pacman domain and a Cartpole domain of up to 125 tasks.