The idea of projects is to get hands-on experience in applying deep reinforcement learning. You can either propose your own project or choose one from here. Project team size can be 1-2 people. We expect more from 2-people project.
The result of a project is
- the code (preferably on GitHub),
- either blog post on Medium or paper on Arxiv,
- presentation on the final day.
By these dates you should be able to answer following questions:
- 5.03.2019 - What is the task/environment? What is observation space? What is action space? What is reward?
- 26.03.2019 - What algorithm you are going to use? Do you use existing codebase or implement it from scratch? What infrastructure you plan to use for training?
- 23.04.2019 - We expect some initial results. What improvements you plan over the initial results?
- 28.05.2019 - Final project presentations.