The exercises at the end of this chapter could likely provide several hours of fun. Try and only complete one or two exercises, as we still need to finish the book:

1. Set up and run the PyramidsIL scene to run online IL.
2. Set up and run the PushBlockIL scene to run online IL.
3. Set up and run the WallJump scene to run with online IL. This requires you to modify the scene.
4. Set up and run the VisualPyramids scene to use offline recording. Record a training session then train an agent.
5. Set up and run the VisualPushBlock scene to use offline recording. Use offline IL to train the agent.
6. Set up the PushBlockIL scene to record an observation demo. Then use this offline training to train multiple agents in the regular PushBlock scene.
7. Set up the PyramidsIL scene to record a demo recording. Then use this for offline training to train multiple agents in the regular Pyramids scene.
8. Train an agent in the VisualHallway scene using any form of learning you like. After training, modify the VisualHallway scene to use different materials on the walls and floor. Changing materials on Unity objects is quite easy. Then, use the technique of swapping model checkpoints as a way of transfer learning the previously trained brain into a new environment.
9. Do exercise eight, but using the VisualPyramids scene. You could also add other objects or blocks in this scene.
10. Do exercise eight, but using the VisualPushBlock scene. Try adding other blocks or other objects that the agent may have to work around.

Just remember that, if you are attempting any of the Transfer Learning exercises, attention to detail is important when matching the complex graphs. In the next section, we summarize what we have covered in this chapter.