Currently I'm working on an educational project (implementation of AlphaZero approach to different types of board games). My biggest concern at the moment is how to encode board before input into the neural network?

For example, how can this be done for Kalah game?

Kalah board has two straight rows of six pits, and two large score-houses. Each pit contains 6 seeds (so there are 72 seeds)

  • $\begingroup$ Knowing how to generate the input to a neural network is more of an art than a set of rules. I would look up different ways which have been proven already (such as chess or go) to determine how you could do it for something as Kalah. And of course you should try different methods and compare them, certainly when you are doing this for educational purposes. $\endgroup$ Commented Feb 13, 2020 at 16:04


You must log in to answer this question.