Reinforcment Learning (RL) techniques should be accessible to anyone with an interest in programming and Artificial Intelligence. Unfortunately, the field is broad, with exponential growth in the number of research papers published each year. This repository seeks to generalize these solutions to RL problem, by creating an API which can be used to build any type of agent using any desired framework, support research through clear, transparent, and extensible implementation, and support fast prototyping of complex experimental systems by an easy to use API and library of algorithms.
This project will be built under Python3.6, and will largely use TensorFlow and Keras for standard Machine Learning tools. Many of the example agents will be based on OpenAI/gym environments with the goal of reproducing results for each of the standard RL tasks available there.
This is just the beginning. Here are the main goals:
- Project Description and Sprint Planning
- Create API Usability Tests
- Upfront Design of the API (Interface Only)
And so it Begins!
