Method puct

mcts_math/nodes/mcts_node.py:47–50

(self)

        self.parent.update_recursive(value, start_node)

    def puct(self) -> float:
        q_value = self.q_value() if self.visit_count() > 0 else 0
        u_value = self.c_puct * self.prior * np.sqrt(self.parent.visit_count()) / (1 + self.visit_count())
        return q_value + u_value