To set up, move this base directory into your Battlecode2015 installation and rename it to teams.
We can use basic semantic versioning to manage our submissions, which I'm also tagging as I go (so far) and storing in the /submissions folder.
Yay!
- Implement some high-level actions to use with deep/reinforcement learning
- Build a deep/reinforcement learning network to figure out the best strategies using those actions
- Starting with implementing a naive version of the CF-version of R-SMART and seeing where that takes us
- It'd be cool to have some sort of scout() action that sends a scout around to gather informations about ennemies and the environment
- Actions can be stored using the adapter pattern
- It doesn't seem possible to use memory outside a set of tournament matches, which would make training our player in scrimmage matches impossible. Would it make sense to evolve two of our own learning players against each other? xD
