Speeding up reinforcement learning by combining attention and agency features

Berkay Demirel, Martí Sánchez-Fibla

When playing video-games we immediately detect which entity we control and we center the attention towards it to focus the learning and reduce its dimensionality. Reinforcement Lea…