This is my implementation of a Q-Learning agent, found in classifierAgents.py, and a Decision-Tree classifier agent, found in mlLearningAgents.py, in the Berkeley Pac-Man environment.
The main branch contains the Decision-Tree classifier agent and respective unit tests, and the master branch contains the Q-Learning agent.
- Python 2.7
- Numpy
In your command line enter:
python pacman.py -p QLearnAgent -x 2000 -n 2010 -l smallGrid
This selects the agent QLearnAgent, to train on 2000 games, and run on an additional 10 games (2010-2000=10), on smallGrid layout.
When you've tried this, check out his performance on testClassic.
python pacman.py -p QLearnAgent -x 2000 -n 2010 -l testClassic