Autonomous and Adaptive System M

This repository is the result of the the project work for the exam.

Files

• impala -> contains the impala CNN architecture
• nature -> contains the Nature CNN architecture
• keras-procgen.py -> TD(0)
• reinforce-procgen.py -> REINFORCE with baseline
• advantage-keras-procgen.py -> A2C

How to run

USE_IMPALA=1 NUM_LEVEL=5 ENV_NAME=chaser GENERATE_GIF=1 python run-keras-procgen.py <weights_file>

How to train

USE_IMPALA=1 NUM_LEVEL=5 ENV_NAME=chaser DONE_REWARD=d python reinforce-procgen.py lr clip