epochs = 1000
lr = 0.1
goal_loss = 0.01
neurons_number = [35, 10]
opt_name = "Adam"
optimizer = Adam(lr=lr, decay=0.0001)
draw_step = 10
verbose = 1
# 2 model and data initializing---------------------------------------------------------
(x_train, y_train), (x_test,
y_test) = dataset5.load_data(train_size=train_size,
show=False)
x_train = np.transpose(
np.append(x_train, np.ones(x_train.size)).reshape(2, x_train.size))
x_test = np.transpose(
np.append(x_test, np.ones(x_test.size)).reshape(2, x_test.size))
model = Sequential()
model.add(Dense(neurons_number[0], input_dim=2, activation='sigmoid'))
model.add(Dense(neurons_number[1], activation='sigmoid'))
model.add(Dense(1, activation='linear'))
# 3 setting stopper---------------------------------------------------------
callbacks = [
import LABS.ZeroLab.E_Function as dataset5
from neupy import algorithms
import matplotlib.pyplot as plt
import numpy as np
if __name__ == '__main__':
(x_train, y_train), (x_test, y_test) = dataset5.load_data(train_size=6000, show=False)
for std in [0.1, 0.01, 0.001]:
nw = algorithms.GRNN(std=std, verbose=True)
nw.train(x_train, y_train)
y_pred = nw.predict(x_test)
mae = (np.abs(y_test - y_pred)).mean()
plt.plot(x_test, y_test, 'b.', label='real')
plt.plot(x_test, y_pred, 'r.', label='fit')
plt.legend(loc='upper right')
plt.title('GRNN aprox with neupy\nstd = %.4f\nmae = %.3f' % (std,mae))
plt.show()
def get_index(poss_code):
for i in range(poss_code.__len__()):
if poss_code[i] == 1:
return i
if __name__ == '__main__':
epochs = 100
step = 0.5
train_size = 4000
goal_loss = 0.01
(x_train, y_train), (x_test,
y_test) = dataset5.load_data(train_size=train_size,
mode=1)
data = zip(x_train, y_train)
for m in [5]:
sofm = algorithms.SOFM(n_inputs=2,
n_outputs=m * m,
step=step,
show_epoch=20,
verbose=True,
learning_radius=1,
features_grid=(m, m, 1),
epoch_end_signal=on_epoch_end)
sofm.train(data, epochs=10)