An implementation of deep Q-learning neural network playing ViZDoom that uses TF Eager Execution 
This project was completed during CodeJam 2018 at McGill University. Here's the Devpost submission.
We've re-implemented a CNN used in the original article using TF Eager Execution and put an agent into a different environment called "Take Cover". See more detailed description on the Devpost page.
- Deep Q learning with Doom article and notebook by @simoninithomas was used heavily as a guide.
- Tensorflow Eager Execution was used to re-write the deep learning model and training procedure in order to make them easier to understand and improve performance.
- ViZDoom Doom-based AI research platform for reinforcement learning from raw visual information. It allows developing AI bots that play Doom using only the screen buffer
- SLADE3 Doom map scenario and script editor. It was used to customize the environment (e.g. speeding up the agent's and enemies' speed reduces the lenght of each training episode and overall training time).