Using Advice to Transfer Knowledge Acquired in One Reinforcement Learning Task to Another