Graceful task adaptation with a bi-hemispheric RL agent

TL;DR

This paper introduces a bi-hemispheric reinforcement learning agent inspired by human brain lateralization, enabling effective adaptation to novel tasks with minimal performance loss and potential for continual learning.

Contribution

We propose a novel RL agent with specialized hemispheres based on the NRH, improving task adaptation and maintaining learning efficiency on new tasks.

Findings

The agent effectively transfers knowledge from the right hemisphere to new tasks.

Minimal performance degradation on novel tasks compared to traditional RL agents.

Potential for extension to continual learning scenarios.

Abstract

In humans, responsibility for performing a task gradually shifts from the right hemisphere to the left. The Novelty-Routine Hypothesis (NRH) states that the right and left hemispheres are used to perform novel and routine tasks respectively, enabling us to learn a diverse range of novel tasks while performing the task capably. Drawing on the NRH, we develop a reinforcement learning agent with specialised hemispheres that can exploit generalist knowledge from the right-hemisphere to avoid poor initial performance on novel tasks. In addition, we find that this design has minimal impact on its ability to learn novel tasks. We conclude by identifying improvements to our agent and exploring potential expansion to the continual learning setting.