Continual and Multi-task Reinforcement Learning With Shared Episodic Memory
This work addresses the challenge of knowledge transfer and skill acquisition in sequential multi-task environments for reinforcement learning agents, presenting an incremental improvement over existing methods.
The paper tackled the problem of improving efficiency in multi-task and continual reinforcement learning by introducing a neural architecture with shared episodic memory (SEM) that separates episodic and task-specific encoding. The result showed that an agent using SEM effectively reused episodic knowledge from other tasks to enhance its policy on the current task in the Taxi problem, facilitating novel skill acquisition in continual learning.
Episodic memory plays an important role in the behavior of animals and humans. It allows the accumulation of information about current state of the environment in a task-agnostic way. This episodic representation can be later accessed by down-stream tasks in order to make their execution more efficient. In this work, we introduce the neural architecture with shared episodic memory (SEM) for learning and the sequential execution of multiple tasks. We explicitly split the encoding of episodic memory and task-specific memory into separate recurrent sub-networks. An agent augmented with SEM was able to effectively reuse episodic knowledge collected during other tasks to improve its policy on a current task in the Taxi problem. Repeated use of episodic representation in continual learning experiments facilitated acquisition of novel skills in the same environment.