发帖
雷达卡
Nearly Optimal Reward-Free Reinforcement Learning
Zihan Zhang 1 Simon S. Du 2 Xiangyang Ji 1
Abstract RL is exploration for which the agent needs to strategically
We study the reward-free reinforcement learn- visit new states to learn transition and reward information
ing framework, which is particularly suitable for therein. To execute efficient exploration, the agent must
batch reinforcement learning and scenarios where foll ...
京公网安备 11010802022788号
