facebook VizDoom论文 facebook 在机器的 VizDoom 比赛中得到了第一名,其中涉及到了强化学习在sparse reward的环境中使用Reward Shaping和Curriculum Learning的技巧。
Connecting Generative Adversarial Network and Actor Critic Methods.pdf Connecting Generative Adversarial Network and Actor-Critic Methods.pdf