Rainbow with Episodic Memory in Deep Reinforcement Learning

Recently, episodic memory based deep reinforcement learning using nonparametric value estimation such as EVA has attracted attention because these methods can improve sample efficiency and converge learning faster. In this paper, we propose a method that combines Rainbow with episodic memory by extending EVA in order for an agent to perform better. We show this method improves an agent performance in terms of both sample efficiency and getting high scores through some experiments on Atari environment.