Hands-On Reinforcement Learning for Games

Exercises

The further we progress in this book, the more valuable and expensive each of these exercises will become. By expensive, we mean the amount of time you need to invest in each will increase. That may mean you are inclined to do fewer exercises, but please continue to try and do two or three exercises on your own:

1. Revisit TensorSpace Playground and see if you can understand the difference pooling makes in those models. Remember that we avoid the use of pooling in order to avoid losing spatial integrity.
2. Open Chapter_7_DQN_CNN.py and alter some of the convolutional layer inputs such as the kernel or stride size. See what effect this has on training.
3. Tune the hyperparameters or create new ones for Chapter_7_DoubleDQN.py.
4. Tune the hyperparameters ...