DCT: Dual Channel Training of Action Embeddings for Reinforcement Learning with Large Discrete Action Spaces

Open in new window