- Create the basic game
- Implement Carta counting algorithm
- Train Deep Learning Model on playing the game
This is a minimal Version of the Go board game. Because the board size is settable, the complexity can be reduced at will. This might make it possible to try out various unsupervised learning models without the need for a Supercomputer.
Currently the point evaluation is beeing adjusted to fit the best approximation when the both players pass. See rules if you are not familiar with Go.
- Improve point evaluation accuracy at game end
- This has shown to be far from trivial
- This paper seems to be the best solution so far and will be implemented
-> See Projects