def train_policies(self, load_best_policy=False, load_reinforcement=False):
if load_reinforcement:
for i in range(0, game_setting.K):
policy = Policy(self.game_setting)
file_name = policy.load_reinforcement_model(i)
self.policies.append([policy, file_name, 0, 0])
return
if load_best_policy:
start = 1
policy = Policy(self.game_setting)
nr_of_training_cases = policy.load_best_model()
self.policies.append([policy, nr_of_training_cases, 0, 0])
else:
start = 0
policy = Policy(self.game_setting)
max_cases = min(
policy.import_data_and_train(max_cases=self.max_cases,
test_nr_of_cases=True),
self.max_cases)
if self.negative_training_power > 0:
for i in range(start, self.K):
nr_of_cases = max(
0, max_cases // ((i + 1)**self.negative_training_power))
if nr_of_cases > 0:
policy = Policy(self.game_setting)
actual_nr_of_cases = policy.import_data_and_train(
max_cases=nr_of_cases)
else:
policy = Policy(self.game_setting, no_model=True)
actual_nr_of_cases = 0
self.policies.append([policy, actual_nr_of_cases, 0, 0])
else:
for i in range(start, self.K):
policy = Policy(self.game_setting)
nr_of_cases = max(
int(max_cases * (self.K - i - 1) / (self.K - 1)), 0)
if nr_of_cases > 0:
actual_nr_of_cases = policy.import_data_and_train(
max_cases=nr_of_cases)
self.policies.append([policy, actual_nr_of_cases, 0, 0])
else:
self.policies.append([policy, 0, 0, 0])
