Building a Subspace of Policies for Scalable Continual Learning
This addresses the challenge of scalable continual learning for autonomous agents, offering a novel method that is incremental in improving upon existing approaches.
The paper tackles the problem of balancing model size and performance in continual learning for autonomous agents by introducing Continual Subspace of Policies (CSP), which builds a subspace of policies that grows sublinearly with tasks, avoids forgetting, and shows positive transfer, outperforming baselines in Brax and Continual World domains.
The ability to continuously acquire new knowledge and skills is crucial for autonomous agents. Existing methods are typically based on either fixed-size models that struggle to learn a large number of diverse behaviors, or growing-size models that scale poorly with the number of tasks. In this work, we aim to strike a better balance between an agent's size and performance by designing a method that grows adaptively depending on the task sequence. We introduce Continual Subspace of Policies (CSP), a new approach that incrementally builds a subspace of policies for training a reinforcement learning agent on a sequence of tasks. The subspace's high expressivity allows CSP to perform well for many different tasks while growing sublinearly with the number of tasks. Our method does not suffer from forgetting and displays positive transfer to new tasks. CSP outperforms a number of popular baselines on a wide range of scenarios from two challenging domains, Brax (locomotion) and Continual World (manipulation).