search

junxiaosong / AlphaZero_Gomoku

An implementation of the AlphaZero algorithm for Gomoku (also called Gobang or Five in a Row)
MIT License
3.34k stars 970 forks source link

TensorFlow input_states有些疑问 #101

Open MaybeChow1991 opened 4 years ago

MaybeChow1991 commented 4 years ago

self.input_states = tf.placeholder( tf.float32, shape=[None, 4, board_height, board_width],name="input_states")

第二的参数为什么是4,代表着什么?按道理不是应该是3(3种情况,玩家1,玩家2,未下),有人可以帮我解答一下吗?

cloxnu commented 4 years ago

1: 当前玩家落子 2: 对方玩家落子 3: 上一步落子 4: 当前玩家，全是 0 or 1

image