def child_finder(self, node, montecarlo): if node.state == 0: node.add_children([Node(1), Node(-1)]) else: for i in range(2): modifier = (100 if i == 1 else 200) * (-1 if node.state < 0 else 1) node.add_child(Node(node.state + modifier))
def add_child_to_parent(self, parent, move):
child = Node(deepcopy(parent.state))
child.state.move(move)
child.player_number = child.state.whose_turn()
parent.add_child(child)
return child
def build_montecarlo(self):
root_node = Node(deepcopy(self.game))
root_node.player_number = self.game.whose_turn()
montecarlo = MonteCarlo(root_node)
montecarlo.child_finder = self.montecarlo_child_finder
return montecarlo
def build_children(self, node): children = [] for i in range(2): child = Node(node.state or (1 if i == 1 else -1)) child.policy_value = .90 if i == 1 else 0.10 children.append(child) node.update_win_value(0) return children
def test_choice_is_correct(self): montecarlo = MonteCarlo(Node(0)) montecarlo.child_finder = self.child_finder montecarlo.simulate(50) chosen_node = montecarlo.make_choice() self.assertIs(chosen_node.state, 1) exploratory_node = montecarlo.make_exploratory_choice() self.assertTrue(chosen_node != None)
def child_finder(node):
for move in node.state.getMoves():
child = Node(deepcopy(node.state))
child.state.move(*move)
node.add_child(child)
def init_board(state, currentPlayer):
self.board = board.Board(state, currentPlayer)
self.root = Node(board.Board(board.getState(), currentPlayer))
santoriniSim.player_number = 2
montecarlo = MonteCarlo(root)
montecarlo.simulate(self.num_sims)