FROMP in reinforcement learning
Use regularized.py in Google Colab, with images.py and get_memorable.py in the files. regularized.py will train a deep q network on games from MinAtar (miniature Atari) and save models to MyDrive/aim_mini/regularized_models. It will automatically delete older versions of the saved model; empty your Drive's trashcan frequently. After the deep q network is trained with and without regularization, use test_model.py to compare them.
YouTube video describing results: youtube.com/watch?v=fxyttf6T5cA