class Tester:
def __init__(self, number_of_gaussian):
self.ng = number_of_gaussian
self.gmm = GMM(number_of_gaussian)
self.comparator_with_impostor = ComparatorWithImpostor(
number_of_gaussian)
self.comparator = Comparator(number_of_gaussian)
self.cross_validator = CrossValidation(0.3)
def initializer(self):
target_frame_sanderson = pd.read_csv('data/feature.csv')
target_frame_sanderson.fillna(0, inplace=True)
y = target_frame_sanderson.iloc[:, 1].values
X = target_frame_sanderson.iloc[:, 2:].values
self.X_train, self.y_train, self.X_test, self.y_test = self.cross_validator.exec(
X, y)
def plotOne(self, data):
l = len(data)
x = np.linspace(1, l, l)
plt.plot(x, data)
plt.show()
def plot(self, executions, index_models=0):
startTime = datetime.datetime.now()
xAxis = np.arange(executions)
execs = [
self.exec_with_time(index_models) for e in range(0, executions)
]
if index_models == 0:
accepts, errors = zip(*execs)
else:
accepts, errors, impostor = zip(*execs)
finishTime = datetime.datetime.now()
duration = finishTime - startTime
print('Duration Total')
print(duration)
fig, ax = plt.subplots()
rects1 = ax.bar(xAxis, accepts, color='b')
rects2 = ax.bar(xAxis, errors, color='r')
ax.set_ylabel('Taxa de acerto')
ax.set_title('GMM')
ax.legend((rects1[0], rects2[0]), ('Acerto', 'Erro'))
plt.show()
def exec_with_time(self, index_models=0):
startTime = datetime.datetime.now()
if index_models == 0: result = self.exec()
elif index_models == 1: result = self.exec_with_impostor()
else: result = self.exec_with_impostor_universal()
finishTime = datetime.datetime.now()
duration = finishTime - startTime
print('Duration parcial: ')
print(duration)
print('\n')
return result
def models_with_bms(self, X_train):
c0, c1, c2 = X_train
orq1 = np.array(c0)
orq2 = np.array(c1)
orq3 = np.array(c2)
impostorOrq1 = np.concatenate([c1, c2])
impostorOrq2 = np.concatenate([c0, c2])
impostorOrq3 = np.concatenate([c0, c1])
return [(self.gmm.model(orq1), self.gmm.model(impostorOrq1)),
(self.gmm.model(orq2), self.gmm.model(impostorOrq2)),
(self.gmm.model(orq3), self.gmm.model(impostorOrq3))]
def models_with_universal(self, X_train):
cs = [np.array(ci) for ci in X_train]
concated = np.concatenate(cs)
universal = self.gmm.model(concated)
models = [self.gmm.model(ci) for ci in cs]
result = [(model, universal) for model in models]
return result
def models(self, X_train):
cs = [np.array(ci) for ci in X_train]
models = [self.gmm.model(ci) for ci in cs]
return models
def exec(self):
models = self.models(self.X_train)
result = self.comparator.sumary(self.X_test, self.y_test, models)
return result
def exec_with_impostor(self):
models = self.models_with_bms(self.X_train)
result = self.comparator_with_impostor.sumary(self.X_test, self.y_test,
models)
return result
def exec_with_impostor_universal(self):
models = self.models_with_universal(self.X_train)
result = self.comparator_with_impostor.sumary(self.X_test, self.y_test,
models)
return result
class Tester:
def __init__(self, number_of_gaussian):
self.ng = number_of_gaussian
self.gmm = GMM(number_of_gaussian)
self.comparator_with_impostor = ComparatorWithImpostor(
number_of_gaussian)
self.comparator = Comparator(number_of_gaussian)
self.cross_validator = CrossValidation(0.2)
def initializer(self):
data, target = load_iris(True)
self.data_frame = []
for d in data:
df_t = pd.DataFrame(d)
self.data_frame.append(df_t)
self.target_frame = pd.DataFrame(target)
self.X_train, self.y_train, self.X_test, self.y_test = self.cross_validator.exec(
data, target)
def plot(self, executions, index_models=0):
startTime = datetime.datetime.now()
xAxis = np.arange(executions)
execs = [
self.exec_with_time(index_models) for e in range(0, executions)
]
if index_models == 0:
accepts, errors = zip(*execs)
else:
accepts, errors, impostor = zip(*execs)
finishTime = datetime.datetime.now()
duration = finishTime - startTime
print('Duration Total')
print(duration)
fig, ax = plt.subplots()
rects1 = ax.bar(xAxis, accepts, color='b')
rects2 = ax.bar(xAxis, errors, color='r')
ax.set_ylabel('Taxa de acerto')
ax.set_title('GMM')
ax.legend((rects1[0], rects2[0]), ('Acerto', 'Erro'))
plt.show()
def exec_with_time(self, index_models=0):
startTime = datetime.datetime.now()
if index_models == 0: result = self.exec()
elif index_models == 1: result = self.exec_with_impostor()
else: result = self.exec_with_impostor_universal()
finishTime = datetime.datetime.now()
duration = finishTime - startTime
print('Duration parcial: ')
print(duration)
print('\n')
return result
def models_with_bms(self, X_train):
c0, c1, c2 = X_train
orq1 = np.array(c0)
orq2 = np.array(c1)
orq3 = np.array(c2)
impostorOrq1 = np.concatenate([c1, c2])
impostorOrq2 = np.concatenate([c0, c2])
impostorOrq3 = np.concatenate([c0, c1])
return [(self.gmm.model(orq1), self.gmm.model(impostorOrq1)),
(self.gmm.model(orq2), self.gmm.model(impostorOrq2)),
(self.gmm.model(orq3), self.gmm.model(impostorOrq3))]
def models_with_universal(self, X_train):
c0, c1, c2 = X_train
orq1 = np.array(c0)
orq2 = np.array(c1)
orq3 = np.array(c2)
impostorUniversal = np.concatenate([c0, c1, c2])
universal = self.gmm.model(impostorUniversal)
return [(self.gmm.model(orq1), universal),
(self.gmm.model(orq2), universal),
(self.gmm.model(orq3), universal)]
def models(self, X_train):
c0, c1, c2 = X_train
orq1 = np.array(c0)
orq2 = np.array(c1)
orq3 = np.array(c2)
return [
self.gmm.model(orq1),
self.gmm.model(orq2),
self.gmm.model(orq3)
]
def exec(self):
models = self.models(self.X_train)
result = self.comparator.sumary(self.X_test, self.y_test, models)
return result
def exec_with_impostor(self):
models = self.models_with_bms(self.X_train)
result = self.comparator_with_impostor.sumary(self.X_test, self.y_test,
models)
return result
def exec_with_impostor_universal(self):
models = self.models_with_universal(self.X_train)
result = self.comparator_with_impostor.sumary(self.X_test, self.y_test,
models)
return result
