tfddqn Learns with Double Q-Learning.

EndGameScore returns weather it's the end of the game, and the hard-coded score of a win/loss/draw for the current player if it is finished.

OneHotEncoding returns a slice of float32 with one element set to 1, and all others to 0.

WinGameScore for the winning side.

BatchBoardScorerWrapper is a trivial implementation of a BatchBoardScorer, with no efficiency gains.

BatchBoardScorer is a BoardScorer that handles batches.

BoardScorer returns two scores: value: how likely the current player is to win, in the form of a score from +10 and -10.

LearnerScorer is the interface used to train a model.

PolicyBatchScorer represents an AI capable of scoring individual actions for each board.

PolicyScorer represents an AI capable of scoring both the board and individual actions.