Most relevant files in this repository:
- Depp_Learning_Report.pdf Final report
- Final_Notebook.ipynb Has example code to start training, loads pre-trained models and runs simple evaluation of selected models.
- Depp_Learning_Poster.pdf Poster used in poster-session.
- Impala_without_death_penalty_reward_10.mp4 and Impala_with_death_penalty_5_reward_62.mp4 in Videos folder. Video examples that illustrate the behavioural differences highlighted in the report.
- all the .py files contains the source code used to create and train models.
Most of the other files required to be mounted on Google Drive and have access to files generated by the authors and will not be able to run as-is for other users.
Reinforcement Learning with Augmented Data: Article, Git
IMPALA: Scalable Distributed Deep-RL with Importance Weighted Actor-Learner Architectures: Article, Git
Leveraging Procedural Generation to Benchmark Reinforcement Learning: Article, Git
Proximal Policy Optimization Algorithms: Article