def prepare(filenames):
X = np.array([])
Y = np.array([])
for each in filenames:
data = pd.read_csv(each)
x = data.iloc[:, :-1].values
y = data.iloc[:, 1].values
X = np.append(X, normalize(x))
Y = np.append(Y, normalize(y))
return X, Y
def _normalize(self, x_train, y_train, x_test, y_test):
if not self.has_normalized:
self.max_x = np.max(x_train)
self.min_x = np.min(x_train)
self.max_y = np.max(y_train)
self.min_y = np.min(y_train)
x_train, y_train = normalize(x_train), normalize(y_train)
if x_test is not None:
x_test = (x_test - self.min_x) / (self.max_x - self.min_x + 1)
if y_test is not None:
y_test = (y_test - self.min_y) / (self.max_y - self.min_y + 1)
return x_train, y_train, x_test, y_test
else:
print("has already normalized...")
def generate_X(filename, sample=None):
'''
take a sample from data
'''
print('Generating X...')
data = pd.read_csv(filename)
X, Y = normalize(data.iloc[:, 0].values), normalize(data.iloc[:, 1].values)
if sample:
idx = np.random.choice(np.arange(len(X)), sample, replace=False)
X = X[idx]
Y = Y[idx]
data = np.zeros((len(X), 2))
data[:, 0] = X
data[:, 1] = Y
return data
def process_X(X, betas):
Xs = X[:, 0]
dists = []
for x in Xs:
dist = np.abs(np.asarray(x) - np.asarray(betas)).min()
dists.append(dist)
dists = normalize(dists)
dists = np.power((1 - dists), 3)
print(dists.max())
Y = np.array(dists)
return Y
def find_locations(self, predicted_betas):
indices = (-predicted_betas).argsort()[:self.num_breaks]
indices = normalize(indices)
return sorted(indices)
def find_locations(predicted_betas, num_breaks):
indices = (-predicted_betas).argsort()[:num_breaks]
indices = normalize(indices)
return indices