You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
In the search() method of the MCTS class, when a leaf node is reached, after the expansion, action prior probability and leaf state value initialization from the NN, shouldn't we set the leaf node visit count (self.Ns[s]) to 1 instead of 0 (line 101 in the MCTS.py here). This visit count isn't incremented during the backpropagation later, so I think it should be set to 1 at initialization.
The text was updated successfully, but these errors were encountered:
In the search() method of the MCTS class, when a leaf node is reached, after the expansion, action prior probability and leaf state value initialization from the NN, shouldn't we set the leaf node visit count (
self.Ns[s]
) to 1 instead of 0 (line 101 in the MCTS.py here). This visit count isn't incremented during the backpropagation later, so I think it should be set to 1 at initialization.The text was updated successfully, but these errors were encountered: