Skip to content

Latest commit

 

History

History
23 lines (19 loc) · 814 Bytes

README.md

File metadata and controls

23 lines (19 loc) · 814 Bytes

Machine Learning Nanodegree capstone - Solving Breakout with Deep Reinforcement Learning

In this capstone project I evaluate the algorithms DQN e A2C to make an AI that learns how to play Breakout.

In this project I used:

  • OpenAI Gym.
  • OpenAI Baselines adapted for the project, in this package. I left this working copy in the state I ran the experiments.
  • TensorFlow
  • cloudpickle
  • numpy
  • pandas
  • scipy
  • joblib

There is an AWS EC2 Instance ready for this and public: ami-1c79d264. This AMI have all needed packages to run the experiment and a shell script to facilitate the work. It is described in the report how to use it.

I could not use IPython notebooks because baselines package was not practical to work with Jupyter notebooks and I had to extract the results from TensorBoard and CSVs.