A Step towards Artificial Pokemon Master

Kaishang CHEN

I introduced a novel multi-agent competition-based framework to train reinforcement learning agents to play Pokemon. Experiments showed that agents trained with out method outperformed those trained with self-play, even without reward shaping.