import random
from bke import MLAgent, is_winner, opponent, RandomAgent, train_and_plot
class MyAgent(MLAgent):
def evaluate(self, board):
if is_winner(board, self.symbol):
reward = 1
elif is_winner(board, opponent[self.symbol]):
reward = -1
else:
reward = 0
return reward
random.seed(1)
my_agent = MyAgent()
random_agent = RandomAgent()
train_and_plot(agent=my_agent,
validation_agent=random_agent,
iterations=50,
trainings=100,
validations=1000)
def train_and_plot(self, iterations, trainings, validations):
random_agent = RandomAgent()
train_and_plot(agent=self, validation_agent=random_agent, iterations=iterations, trainings=trainings, validations=validations)
save(self, 'MyAgent')
class MyAgent(MLAgent):
def evaluate(self, board):
if is_winner(board, self.symbol):
reward = 1
elif is_winner(board, opponent[self.symbol]):
reward = -1
else:
reward = 0
return reward
random.seed(1)
my_agent = MyAgent(alpha=0.2, epsilon=0.5)
my_agent.learning = False
random_agent = RandomAgent()
train_and_plot(agent=my_agent,
validation_agent=random_agent,
iterations=50,
trainings=100,
validations=1000)
train(my_agent, 3000)
validation_agent = RandomAgent()
validation_result = validate(agent_x=my_agent,
agent_o=validation_agent,
iterations=100)
def validate(self):
self.learning = False
validation_agent = RandomAgent()
validation_result = validate(agent_x=self, agent_o=validation_agent, iterations=1000)
plot_validation(validation_result)