Towards sample-efficient episodic control with DAC-ML
This work aims to improve the sample efficiency of reinforcement learning models, which is a problem for researchers developing AI agents that can learn behavioral policies with fewer episodes.
This paper addresses the sample-inefficiency problem in Deep Reinforcement Learning by introducing DAC-ML, a novel cognitive architecture inspired by the Distributed Adaptive Control theory. DAC-ML, which includes a hippocampus-inspired sequential memory system, rapidly converges to effective action policies for reward maximization in a challenging foraging task.
The sample-inefficiency problem in Artificial Intelligence refers to the inability of current Deep Reinforcement Learning models to optimize action policies within a small number of episodes. Recent studies have tried to overcome this limitation by adding memory systems and architectural biases to improve learning speed, such as in Episodic Reinforcement Learning. However, despite achieving incremental improvements, their performance is still not comparable to how humans learn behavioral policies. In this paper, we capitalize on the design principles of the Distributed Adaptive Control (DAC) theory of mind and brain to build a novel cognitive architecture (DAC-ML) that, by incorporating a hippocampus-inspired sequential memory system, can rapidly converge to effective action policies that maximize reward acquisition in a challenging foraging task.