Policy network: 12 layer convolutional neural network. Training data: 30M positions from human expert games (KGS 5+ dan).
確定! 回上一頁