My-Research-Project-2022

Honours Research Project Title: Surrogate Tasks for Architecture Selection in Reinforcement Learning

Investigating the correlation between supervised learning and deep reinforcement learning networks regarding the performance of the agent.

Phase 1

needs data from Atari HEAD :

@dataset{ruohan_zhang_2019_3451402,

author = {Ruohan Zhang and Calen Walshe and Zhuode Liu and Lin Guan and Karl S. Muller and Jake A. Whritner and Luxin Zhang and Mary Hayhoe and Dana Ballard},

title = {{Atari-HEAD: Atari Human Eye-Tracking and Demonstration Dataset}},

month = sep,

year = 2019,

publisher = {Zenodo},

version = 4,

doi = {10.5281/zenodo.3451402},

url = {https://doi.org/10.5281/zenodo.3451402}

}

Games used:

*Breakout

*Frostbite

*Enduro

sh files

SL.sh runs SL_python.py -> sbatch SL.sh

DeepRL.sh runs DeepRL.py -> sbatch DeepRL.sh

new_script.sh has all the necessary commands to install libraries using anaconda.

AI data

Notebook used to generate the data is found in the AI_generated_data folder -> has link to google colab

Google drive link to data: https://drive.google.com/drive/folders/12EZlR3KVwsMhojgnc1Rpf8HR-6nY7b2o?usp=sharing

Structure of Repo:

.
├── AI                  
│   ├── DQN.ipynb         # Code for DQN agent 
│   ├── videos            # Folder with best DQN runs videos
├── Phase 1         
│   ├── A2C.ipynb         # Code for Actor Critic agent 
│   ├── videos            # Folder with best A2C runs videos
├── Report.pdf            # Report 
└── README.md