This is a weird thing. The PCA result, as expected, is basically what the moves of chess for the major pieces look like. Also, I don't think they understand wha

This is a weird thing. The PCA result, as expected, is basically what the moves of chess for the major pieces look like. Also, I don't think they understand what bitboards are for (move generation and evaluation in a fairly dense/high performance representation); this vector representation is just serial 1-hot encoding. There are alternate vector encodings which actually make more sense for a neural net or matrix representation of chess positions. I guess the authors have a company that does some kind of NNMF thing on genetic data, which would explain their approach.