Natural Gradient Deep Q-Learning (NGDQN)

Please find our paper on arXiv here.

This repository contains supplementary code for natural gradient Q-Learning. As of right now, the repository just contains the baseline code, which you can run yourself to replace the baseline results.

Baselines

The final scripts from the grid search used for crossvalidation located in baselines/cross_validate/py/

Cross validation can be run with python cross_validate.py, and results can be viewed with python cross_validate_results.py. To run the grid search, run python parallel_slurm.py.