Parham Nooranbakht

Decoupled Q-learning achieves state-of-the-art performance on continuous control by adaptively growing action grids, but surprisingly, the adaptive growth itself is not the key driver of success.