Near-Optimal Randomized Exploration for Tabular Markov Decision Processes

We study algorithms using randomized value functions for exploration in reinforcement learning. This type of algorithms enjoys appealing empirical performance.