Project Group 9
This repository contains group 9's code for our math 156 machine learning group project.
See our project_proposal.pdf for more details about our model.
This folder contains lots of jupyter notebook files used for testing, but it was not a robust method for version control with many people. Note: not all files run properly in this folder as most were used for our initial testing.
This folder contains the .py files that were used for training the deep Q networks.
Each test that was trained to completion creates a data folder with relevant logs, plots, and model objects.
- DQN.py defines the deep q learning class
- MC_v3.py is the main file that trains the models
- functions.py has the various reward functions and other auxiliary functions
- jon_train.py is Jonathn's slightly modified train file
- oscar_param_dicts.py holds hyperparameter variations to use while training
- plot_progress.py creates the score, step, and phase plots given a model
- record_model.py creates a video of the mountain car from the resulting model
- shawn_param.py hold hyperparameter variations to use while training
- train.py has the code for iterating through steps and updating the necessary values
- visualizations.ipynb creates the decision space plots for a given model
Our final project report. Describes motivation, theory, methods, and results of the project.