Transfer learning between RTS combat scenarios using component-action deep reinforcement learning Conferences