Target-Based Temporal-Difference Learning

Donghwan Lee, Niao He

Abstract: The use of target networks has been a popular and key component of recent deep Q-learning algorithms for reinforcement learning, yet little is known from the theory side. In this work, we introduce...