def validate(zone, start, end, phase):
#1-35////////35-41
data_validation = feature_extractor.extractor(zone, start, end, phase)
validation_data = Orange.data.Table(
os.path.join(package_directory, 'learner/data_' + zone + '.tab'))
classifier = pickle.load(
open(os.path.join(package_directory, 'learner/ANN_' + zone)))
# data_validation = Orange.data.Table('data_'+zone+'_validate.tab')
i = 1
j = 1
for e in data_validation:
i = i + 1
# print e
# print e.get_class()
# print classifier(e, Orange.classification.Classifier.GetBoth)
# print classifier(e, Orange.classification.Classifier.GetValue)
if classifier(
e, Orange.classification.Classifier.GetValue) == e.get_class():
j = j + 1
# print e.get_class()
print "############" + zone + "############"
print "Classification Accuracy (" + zone + "):" + str(float(j / i))
print "i " + str(i)
print "j " + str(j)
print "################################"
def validate(zone, start, end, phase):
#1-35////////35-41
data_validation= feature_extractor.extractor(zone, start,end, phase)
validation_data = Orange.data.Table('learner/data_'+zone+'.tab')
classifier = pickle.load(open('learner/ANN_'+zone))
# data_validation = Orange.data.Table('data_'+zone+'_validate.tab')
i=1
j=1
for e in data_validation:
i=i+1
# print e
# print e.get_class()
# print classifier(e, Orange.classification.Classifier.GetBoth)
# print classifier(e, Orange.classification.Classifier.GetValue)
if classifier(e, Orange.classification.Classifier.GetValue)==e.get_class():
j=j+1
# print e.get_class()
print "############"+zone+"############"
print "Classification Accuracy ("+zone+"):"+str(float(j/i))
print "i "+str(i)
print "j "+str(j)
print "################################"
def get_best_action(self, strategy, player):
"""
Get the best action for the given player based on given strategy
:return: the best action and the feature of this action
"""
actions = self.game.get_actions(player)
action = None
if not actions:
action = (player, None)
elif strategy == "q":
action = actions[np.argmax(
[self.weights @ extractor(self.game, a) for a in actions])]
elif strategy == "random":
action = actions[random.randint(0, len(actions) - 1)]
feature = extractor(self.game.copy(), action)
return feature, action
def train(zone, start, end, phase, learner, n_mid, reg_fact, max_iter, normalize,rand ):
#1-35////////35-41
data_training= feature_extractor.extractor(zone, start,end, phase)
if learner == True:
# learner = Orange.classification.neural.NeuralNetworkLearner(data_training,n_mid=n_mid, reg_fact=reg_fact, max_iter=max_iter, normalize=normalize, rand=rand)
learner = NN.NeuralNetworkLearner(data_training,n_mid=n_mid, reg_fact=reg_fact, max_iter=max_iter, normalize=normalize, rand=rand)
pickle.dump(learner, open('learner/ANN_'+zone, 'w'))
# get the classifier names for later
cross_val = Orange.evaluation.testing.cross_validation([Orange.classification.neural.NeuralNetworkLearner(n_mid=1, reg_fact=1, max_iter=1, normalize=True, rand=None)], data_training,5)
cross_val_data = Orange.evaluation.testing.ExperimentResults(
[cross_val.classifier_names[0]], cross_val.class_values,
)
# print cross_val_data.classifier_names
fileIO.write_label_file(cross_val_data.classifier_names, 'learner/'+zone+'ZoneLabels')
def train(zone, start, end, phase, learner, n_mid, reg_fact, max_iter,
normalize, rand):
#1-35////////35-41
data_training = feature_extractor.extractor(zone, start, end, phase)
if learner == True:
# learner = Orange.classification.neural.NeuralNetworkLearner(data_training,n_mid=n_mid, reg_fact=reg_fact, max_iter=max_iter, normalize=normalize, rand=rand)
learner = NN.NeuralNetworkLearner(data_training,
n_mid=n_mid,
reg_fact=reg_fact,
max_iter=max_iter,
normalize=normalize,
rand=rand)
fname = os.path.join(package_directory, 'learner/ANN_' + zone)
# with open(fname, 'rb') as file_:
# pickle.dump(learner, open(file_)
with open(fname, 'w') as f:
pickle.dump(learner, f)
# print os.path.realpath(os.path.dirname(__file__))
# print os.path.abspath(__file__)
# pickle.dump(learner, open(package_directory+'/learner/ANN_'+zone, 'w'))
#open(os.path.join(package_directory,'learner/ANN_lower'))
# get the classifier names for later
cross_val = Orange.evaluation.testing.cross_validation([
Orange.classification.neural.NeuralNetworkLearner(
n_mid=1, reg_fact=1, max_iter=1, normalize=True, rand=None)
], data_training, 5)
cross_val_data = Orange.evaluation.testing.ExperimentResults(
[cross_val.classifier_names[0]],
cross_val.class_values,
)
# print cross_val_data.classifier_names
fileIO.write_label_file(cross_val_data.classifier_names,
'learner/' + zone + 'ZoneLabels')