def test_handle_errors(self):
with self.assertRaises(ValueError):
# size of target data not the same as
# size of input data.
pnnet = algorithms.PNN(verbose=False)
pnnet.train(np.array([[0], [0]]), np.array([0]))
with self.assertRaises(ValueError):
# 2-D target vector (must be 1-D)
pnnet = algorithms.PNN(verbose=False)
pnnet.train(np.array([[0]]), np.array([[0, 0]]))
with self.assertRaises(ValueError):
# invalid feature size for prediction data
pnnet = algorithms.PNN(verbose=False)
pnnet.train(np.array([[0], [0]]), np.array([0]))
pnnet.predict(np.array([[0]]))
msg = "hasn't been trained"
with self.assertRaisesRegexp(NotTrained, msg):
# predict without training
pnnet = algorithms.PNN(verbose=False)
pnnet.predict(np.array([[0]]))
with self.assertRaises(ValueError):
# different number of features for
# train and test data
grnet = algorithms.PNN(verbose=False)
grnet.train(np.array([[0]]), np.array([0]))
grnet.predict(np.array([[0, 0]]))
def test_predict_probability(self):
dataset = datasets.load_digits()
x_train, x_test, y_train, y_test = train_test_split(
dataset.data, dataset.target, train_size=0.7
)
x_train_before = x_train.copy()
x_test_before = x_test.copy()
y_train_before = y_train.copy()
number_of_classes = len(np.unique(dataset.target))
pnnet = algorithms.PNN(verbose=False, std=10)
pnnet.train(x_train, y_train)
result = pnnet.predict_proba(x_test)
n_test_inputs = x_test.shape[0]
self.assertEqual(result.shape, (n_test_inputs, number_of_classes))
total_classes_prob = np.round(result.sum(axis=1), 10)
np.testing.assert_array_equal(
total_classes_prob,
np.ones(n_test_inputs)
)
old_result = result.copy()
# Test problem with variable links
np.testing.assert_array_equal(x_train, x_train_before)
np.testing.assert_array_equal(x_test, x_test_before)
np.testing.assert_array_equal(y_train, y_train_before)
x_train[:, :] = 0
result = pnnet.predict_proba(x_test)
total_classes_prob = np.round(result.sum(axis=1), 10)
np.testing.assert_array_almost_equal(result, old_result)
def test_basic_storage(self):
input_data = np.random.random((100, 2))
target_data = np.random.random(100) > 0.5
pnn = algorithms.PNN(std=0.123, verbose=True)
pnn.train(input_data, target_data)
stored_pnn = pickle.dumps(pnn)
loaded_pnn = pickle.loads(stored_pnn)
testcases = [
('pnn', pnn),
('loaded_pnn', loaded_pnn),
]
for name, network in testcases:
print("Test case name: {}".format(name))
self.assertAlmostEqual(network.std, 0.123)
self.assertAlmostEqual(network.verbose, True)
with catch_stdout() as out:
network.logs.stdout = out
network.logs.write("Test message")
terminal_output = out.getvalue()
self.assertIn("Test message", terminal_output)
pnn_prediction = pnn.predict(input_data)
loaded_pnn_prediction = loaded_pnn.predict(input_data)
np.testing.assert_array_almost_equal(loaded_pnn_prediction,
pnn_prediction)
def test_with_noise():
(x_train, y_train), (x_test, y_test) = dataset8.load_data(mode=0)
pnn = algorithms.PNN(std=1, batch_size=128, verbose=True)
pnn.train(x_train[0:10000], y_train[0:10000])
y_predicted = pnn.predict(x_test)
local_path = 'MY_IMAGES_GIMP/'
for nTest in np.arange(0, 10, 1):
# convert to numpy array
x = get_pxs(local_path + str(nTest) + '.png')
# Inverting and normalizing image
x = 255 - x
x /= 255
x = np.expand_dims(x, axis=0)
x.shape = (1, 784)
prediction = pnn.predict(x)
print('REAL \t\tPREDICTED')
print(str(nTest) + '\t\t\t' + str(prediction[0]))
print("accuracy = %.2f" % (metrics.accuracy_score(y_predicted, y_test)))
for i in range(0, 10000):
x_test[i] = noise(x_test[i], 500)
y_predicted = pnn.predict(x_test)
print("accuracy on noise data = %.2f" %
(metrics.accuracy_score(y_predicted, y_test)))
def __init__(self, std=10):
"""
:param std: float
standard deviation for PDF function, default to 0.1.
:return:
"""
self.pnn = algorithms.PNN(std=std, verbose=False)
def clasification():
df = pd.DataFrame()
for name, adress, price in db.session.query(Place.place_name, Place.place_adress, Place.place_price):
print(name, adress, price)
df = df.append({"name": name, "adress": adress, "price": price}, ignore_index=True)
# db.session.close()
X = pd.get_dummies(data=df[['name', 'adress']])
mean_price = df['price'].mean()
df.loc[df['price'] < mean_price, 'quality'] = 0
df.loc[df['price'] >= mean_price, 'quality'] = 1
print(df)
print(X)
pnn = algorithms.PNN(std=10, verbose=False)
pnn.train(X, df['quality'])
test_str = ['club', 'Ковальський провулок']
count_columns = len(X.columns)
test_list = np.array([0] * count_columns)
test_list[0] = 1
test_list[-2] = 1
test_list = np.reshape(test_list, (1, len(test_list)))
print(test_list)
y_predicted = pnn.predict(test_list)
result = "Ні"
if y_predicted - 1 < 0.0000000000001:
result = "Так"
return render_template('clasification.html', y_predicted=result, test_data=test_list[0], test_str=test_str)
def clasification():
df = pd.DataFrame()
for name, company, salary in db.session.query(OrmVacancy.vacancy_name, OrmVacancy.vacancy_company, OrmVacancy.vacancy_salary):
print(name, company, salary)
df = df.append({"name": name, "company": company, "salary": salary}, ignore_index=True)
mean_p = df['salary'].mean()
df.loc[df['salary'] < mean_p, 'quality'] = 0
df.loc[df['salary'] >= mean_p, 'quality'] = 1
X = pd.get_dummies(data=df[['name', 'company']])
print(df)
print(X)
pnn = algorithms.PNN(std=10, verbose=False)
pnn.train(X, df['quality'])
count_columns = len(X.columns)
string_test = ['frontend', 'google']
test_list = [0] * count_columns
test_list[0] = 1
test_list[-1] = 1
print(test_list)
y_predicted = pnn.predict([test_list])
result = "Ні"
if y_predicted - 1 < 0.0000001:
result = "Так"
return render_template('clasification.html', y_predicted=result, test_data=test_list, string_test=string_test)
def fromPN(cls, pos_train, neg_train, z_trains, y_train, models=None):
'''
a neural network will be trained with z_trains, y_train
Arguments:
pos_train, neg_train same in super class
z_trains {List[DateFrame]}
y_train {Array}
Returns:
ZeroOneSemiNaiveBayesClassifier
'''
sbc = super(ZeroOneHemiNaiveBayesClassifier,
cls).fromPN(pos_train, neg_train)
if models is None:
sbc.models = [
algorithms.PNN(std=np.std(z_train.values), verbose=False)
for z_train in z_trains
]
else:
sbc.models = [
copy.copy(model_dict[model])
if isinstance(model, str) else model for model in models
]
sbc.features2 = [z_train.columns for z_train in z_trains]
sbc.fit(z_trains, y_train)
return sbc
def test_basic_storage(self):
X = np.random.random((100, 2))
y = np.random.random(100) > 0.5
# We keep verbose=True in order to see if value will
# be True when we restore it from the pickle object.
pnn = algorithms.PNN(std=0.123, verbose=True)
pnn.train(X, y)
stored_pnn = pickle.dumps(pnn)
loaded_pnn = pickle.loads(stored_pnn)
testcases = [
('pnn', pnn),
('loaded_pnn', loaded_pnn),
]
for name, network in testcases:
print("Test case name: {}".format(name))
self.assertAlmostEqual(network.std, 0.123)
self.assertAlmostEqual(network.verbose, True)
with catch_stdout() as out:
network.logs.stdout = out
network.logs.write("Test message")
terminal_output = out.getvalue()
self.assertIn("Test message", terminal_output)
pnn_prediction = pnn.predict(X)
loaded_pnn_prediction = loaded_pnn.predict(X)
np.testing.assert_array_almost_equal(loaded_pnn_prediction,
pnn_prediction)
def clasification():
df = pd.DataFrame()
for file_text, rating in db.session.query(ormFiles.file_text,
ormFiles.rating):
print(file_text, rating)
df = df.append({
"file_name": file_text,
"rating": float(rating)
},
ignore_index=True)
# db.session.close()
df['count_symbols'] = df['file_name'].apply(len)
df.loc[df['rating'] < 0.33, 'quality'] = 0
df.loc[df['rating'] >= 0.33, 'quality'] = 1
print(df)
pnn = algorithms.PNN(std=10, verbose=False)
pnn.train(df['count_symbols'], df['quality'])
test_data = 'ewij weioh uia guu aweg'
t_test_data = len(test_data)
y_predicted = pnn.predict([t_test_data])
result = "Ні"
if y_predicted - 1 < 0.0000000000001:
result = "Так"
return render_template('clasification.html',
y_predicted=result,
test_data=test_data)
def test_cutting_exceptions(self):
with self.assertRaises(ValueError):
surgery.cut(algorithms.PNN(), 0, 1)
with self.assertRaises(ValueError):
surgery.cut(self.network, 0, 10)
with self.assertRaises(ValueError):
surgery.cut(self.network, 0, 0)
def PNN(X_train, X_test, y_train, y_test, X_dummy):
environment.reproducible()
pnn = algorithms.PNN(std=0.1, verbose=False)
pnn.train(X_train, y_train)
#print 'done trainin'
y_predicted = pnn.predict(X_test)
y_dummy = pnn.predict(X_dummy)
#print y_predicted
return y_dummy, y_predicted, metrics.accuracy_score(y_test, y_predicted)
def test_predict_different_inputs(self):
pnnet = algorithms.PNN(verbose=False)
data = np.array([[1, 2, 3]]).T
target = np.array([[1, 0, 1]]).T
pnnet.train(data, target)
self.assertInvalidVectorPred(pnnet, data.ravel(), target.ravel(),
decimal=2)
def pnnTrainTestNoNorm(X, Y):
X_train, X_test, Y_train, Y_test = train_test_split(X, Y, train_size=0.7)
pnn = algorithms.PNN(std=10, verbose=False)
pnn.train(X_train, Y_train)
y_predicted = pnn.predict(X_test)
score = metrics.accuracy_score(Y_test, y_predicted)
print("PNN score: ", score)
return score, y_predicted
def test_pnn_non_trivial_class_names_as_strings(self):
# Issue #177: https://github.com/itdxer/neupy/issues/177
x = np.array([10] * 10 + [20] * 10 + [30] * 10)
y = np.array(['cat'] * 10 + ['dog'] * 10 + ['horse'] * 10)
pnn = algorithms.PNN(std=1)
pnn.train(x, y)
y_predicted = pnn.predict(x)
np.testing.assert_array_equal(y, y_predicted)
self.assertEqual(sorted(pnn.classes), ['cat', 'dog', 'horse'])
def test0():
"""
daemon
:return:
"""
dataset = datasets.load_digits()
x_train, x_test, y_train, y_test = train_test_split(dataset.data, dataset.target, train_size=0.7)
nw = algorithms.PNN(std=10, verbose=False)
nw.train(x_train, y_train)
result = nw.predict(x_test)
print metrics.accuracy_score(y_test, result)
def test_pnn_non_trivial_class_names(self):
# Issue #177: https://github.com/itdxer/neupy/issues/177
x = np.array([10] * 10 + [20] * 10 + [30] * 10)
y = np.array([1] * 10 + [2] * 10 + [3] * 10)
pnn = algorithms.PNN(std=1)
pnn.train(x, y)
y_predicted = pnn.predict(x)
np.testing.assert_array_almost_equal(y, y_predicted)
self.assertEqual(sorted(pnn.classes), [1, 2, 3])
def test_digit_prediction(self):
dataset = datasets.load_digits()
x_train, x_test, y_train, y_test = train_test_split(
dataset.data, dataset.target, train_size=0.7
)
nw = algorithms.PNN(verbose=False, std=10)
nw.train(x_train, y_train)
result = nw.predict(x_test)
self.assertAlmostEqual(metrics.accuracy_score(y_test, result),
0.9889, places=4)
def PNN(self):
X, y = self.prepareData()
X_train, X_test, y_train, y_test = train_test_split(X,
y,
test_size=0.2,
random_state=1)
X_train = np.array(X_train, dtype=np.float32)
X_test = np.array(X_test, dtype=np.float32)
y_test = np.array(y_test, dtype=np.int32)
pnn = algorithms.PNN(std=10, verbose=False)
pnn.train(X_train, y_train)
y_predict = pnn.predict(X_test)
return y_predict, y_test
def test_handle_errors(self):
with self.assertRaises(ValueError):
# Wrong: size of target data not the same as size of
# input data.
algorithms.PNN(verbose=False).train(np.array([[0], [0]]),
np.array([0]))
with self.assertRaises(ValueError):
# Wrong: 2-D target vector (must be 1-D)
algorithms.PNN(verbose=False).train(np.array([[0], [0]]),
np.array([[0]]))
with self.assertRaises(AttributeError):
# Wrong: can't use iterative learning process for this
# algorithm
algorithms.PNN(verbose=False).train_epoch()
with self.assertRaises(ValueError):
# Wrong: invalid feature size for prediction data
grnet = algorithms.PNN(verbose=False)
grnet.train(np.array([[0], [0]]), np.array([0]))
grnet.predict(np.array([[0]]))
def test1():
"""
数字型类别
:return:
"""
dataset = datasets.load_digits()
n_samples = dataset.data.shape[0]
ratio = 0.7
x_train, y_train = dataset.data[0:n_samples * ratio, :], dataset.target[0:n_samples * ratio]
x_test, y_test = dataset.data[n_samples * ratio:, :], dataset.target[n_samples * ratio:]
nw = algorithms.PNN(std=10, verbose=False)
nw.train(x_train, y_train)
result = nw.predict(x_test)
print metrics.accuracy_score(y_test, result)
def pnnTrainTestNorm(X, Y):
X_train, X_test, Y_train, Y_test = train_test_split(X, Y, train_size=0.7)
scalar = StandardScaler()
scalar.fit(X_train)
X_train_n = scalar.transform(X_train)
X_test_n = scalar.transform(X_test)
pnn = algorithms.PNN(std=10, verbose=False)
pnn.train(X_train_n, Y_train)
y_predicted = pnn.predict(X_test_n)
score = metrics.accuracy_score(Y_test, y_predicted)
print("PNN score: ", score)
return score, y_predicted
def diff_train():
maes = [0, 0, 0, 0]
train_size = [24000, 12000, 5000, 2000]
std = [0.00035, 0.001, 0.0035, 0.01]
for j in range(0, 4):
(x_train, y_train), (x_test, y_test) = dataset3.load_data(train_size=train_size[j], show=False)
pnn = algorithms.PNN(std=std[j], verbose=True)
pnn.train(x_train, y_train)
y_predicted = pnn.predict(x_test)
mae = (np.abs(y_test - y_predicted)).mean()
plt_x_zero = np.empty(0)
plt_y_zero = np.empty(0)
plt_x_one = np.empty(0)
plt_y_one = np.empty(0)
acc = 0.0
i = 0
for coord in x_test:
if y_predicted[i] < 0.5:
plt_x_zero = np.append(plt_x_zero, coord[0])
plt_y_zero = np.append(plt_y_zero, coord[1])
elif y_predicted[i] >= 0.5:
plt_x_one = np.append(plt_x_one, coord[0])
plt_y_one = np.append(plt_y_one, coord[1])
i += 1
plt.plot(plt_x_zero, plt_y_zero, '.')
plt.plot(plt_x_one, plt_y_one, '.')
plt.title('2 class classification\ntrain size = %d\nstd = %.4f\nmae =%.4f' % (train_size[j], std[j], mae))
maes[j] = mae
plt.xlim(0, 1.3)
plt.ylim(0, 1)
plt.legend(('0 class', '1 class'), loc='upper right', shadow=True)
plt.show()
plt.close()
return train_size, std, maes
def diff_train_size():
train_size = [15000, 10000, 5000, 2000]
std = [2, 1, 0.5, 0.25]
for j in range(0, 4):
(x_train, y_train), (x_test, y_test) = dataset8.load_data(mode=0)
pnn = algorithms.PNN(std=std[j], batch_size=128, verbose=True)
pnn.train(x_train[0:train_size[j]], y_train[0:train_size[j]])
y_predicted = pnn.predict(x_test)
maes = metrics.accuracy_score(y_predicted, y_test)
print("accuracy = %.2f" % (maes))
def PNN(X, y):
X_train, X_test, y_train, y_test = train_test_split(X,
y,
test_size=0.2,
random_state=1)
X_train = np.array(X_train, dtype=np.float32)
X_test = np.array(X_test, dtype=np.float32)
y_test = np.array(y_test, dtype=np.int32)
sc = StandardScaler()
sc.fit(X_train)
X_train = sc.transform(X_train)
X_test = sc.transform(X_test)
pnn = algorithms.PNN(std=10, verbose=False)
pnn.train(X_train, y_train)
y_predict = pnn.predict(X_test)
return y_predict, y_test
def test_pnn_mini_batches(self):
dataset = datasets.load_digits()
n_classes = len(np.unique(dataset.target))
x_train, x_test, y_train, y_test = train_test_split(
dataset.data, dataset.target, train_size=0.7
)
pnnet = algorithms.PNN(verbose=False, batch_size=100)
pnnet.train(x_train, y_train)
y_predicted = pnnet.predict(x_test)
self.assertEqual(y_predicted.shape, y_test.shape)
y_predicted = pnnet.predict_proba(x_test)
self.assertEqual(y_predicted.shape,
(y_test.shape[0], n_classes))
def fromPN(cls, pos_train, neg_train, z_trains, y_train, models=None):
'''
a neural network will be trained with z_trains, y_train
Arguments:
pos_train, neg_train same in super class
z_trains {List[DateFrame]}
y_train {Array}
Returns:
ZeroOneSemiNaiveBayesClassifier
'''
sbc = super(ZeroOneHemiNaiveBayesClassifier,
cls).fromPN(pos_train, neg_train)
if models is None or models == 'grnn':
sbc.models = [
algorithms.GRNN(std=np.std([
a for a in z_train.values.ravel()
if str(a) != 'nan' and a != 0
]),
verbose=False) for z_train in z_trains
]
elif models == 'pnn':
sbc.models = [
algorithms.PNN(std=np.std([
a for a in z_train.values.ravel()
if str(a) != 'nan' and a != 0
]),
verbose=False) for z_train in z_trains
]
elif models == 'svm':
sbc.models = [svm.SVC(kernel='rbf') for z_train in z_trains]
elif models == 'lasso':
sbc.models = [LassoLars() for z_train in z_trains]
else:
sbc.models = [
copy.deepcopy(model_dict[model])
if isinstance(model, str) else copy.deepcopy(model)
for model in models
]
sbc.features2 = [z_train.columns for z_train in z_trains]
sbc.fit(z_trains, y_train)
return sbc
def test_simple_pnn(self):
dataset = datasets.load_iris()
data = dataset.data
target = dataset.target
test_data_size = 10
skfold = StratifiedKFold(n_splits=test_data_size)
avarage_result = 0
for train, test in skfold.split(data, target):
x_train, x_test = data[train], data[test]
y_train, y_test = target[train], target[test]
nw = algorithms.PNN(verbose=False, std=0.1)
nw.train(x_train, y_train)
result = nw.predict(x_test)
avarage_result += sum(y_test == result)
self.assertEqual(avarage_result / test_data_size, 14.4)
def diff_std():
(x_train, y_train), (x_test, y_test) = dataset3.load_data(train_size=12000, show=True)
titles = ["\n\nspread greater than necessary", "\n\nspread optimal", "\n\nspread less than necessary"]
spreads = [0.1, 0.001, 0.0001]
for spread, title in zip(spreads, titles):
pnn = algorithms.PNN(std=spread, verbose=True)
pnn.train(x_train, y_train)
y_predicted = pnn.predict(x_test)
mae = (np.abs(y_test - y_predicted)).mean()
plt_x_zero = np.empty(0)
plt_y_zero = np.empty(0)
plt_x_one = np.empty(0)
plt_y_one = np.empty(0)
i = 0
for coord in x_test:
if y_predicted[i] < 0.5:
plt_x_zero = np.append(plt_x_zero, coord[0])
plt_y_zero = np.append(plt_y_zero, coord[1])
elif y_predicted[i] >= 0.5:
plt_x_one = np.append(plt_x_one, coord[0])
plt_y_one = np.append(plt_y_one, coord[1])
i += 1
plt.plot(plt_x_zero, plt_y_zero, '.')
plt.plot(plt_x_one, plt_y_one, '.')
plt.title(title+'\n2 class classification\nstd = %.4f\nmae =%.4f' % (spread, mae))
plt.xlim(0, 1.3)
plt.ylim(0, 1)
plt.legend(('0 class', '1 class'), loc='upper right', shadow=True)
plt.show()
plt.close()
def clasification():
df = pd.DataFrame()
for name, model, price in db.session.query(Products.product_name, Products.product_model, Recommendation.recommendation_price)\
.join(Recommendation, Recommendation.recommendation_id == Products.recommendation_id):
print(name, model, price)
df = df.append({
"name": name,
"model": model,
"price": float(price)
},
ignore_index=True)
# db.session.close()
mean_p = df['price'].mean()
df.loc[df['price'] < mean_p, 'quality'] = 0
df.loc[df['price'] >= mean_p, 'quality'] = 1
X = pd.get_dummies(data=df[['name', 'model']])
print(df)
print(X)
pnn = algorithms.PNN(std=10, verbose=False)
pnn.train(X, df['quality'])
test_str = ['BMW', 'X5']
count_columns = len(X.columns)
test_list = [0] * count_columns
test_list[1] = 1
test_list[-1] = 1
print(test_list)
y_predicted = pnn.predict([test_list])
result = "Ні"
if y_predicted - 1 < 0.0000001:
result = "Так"
return render_template('clasification.html',
y_predicted=result,
test_data=test_list,
test_str=test_str)