# coding: utf-8 import numpy as np from materials import double_moon from connections.models import Model from connections.layers import DenseLayer if __name__ == '__main__': m = Model() m.add_layer(DenseLayer(2, 5, activation='sigmoid')) m.add_layer(DenseLayer(5, 1, activation='sigmoid')) m.compile(lr=0.1) double_moon = double_moon(50, 80, 1) train_x = double_moon[['x', 'y']].as_matrix() train_y = double_moon['class'].as_matrix() train_y = train_y.reshape((train_y.shape[0], 1)) m.train(train_x, train_y)
# coding: utf-8 import numpy as np from connections.models import Model from connections.layers import DenseLayer, RecurrentLayer if __name__ == '__main__': m = Model() m.add_layer(RecurrentLayer(2, 5, activation='tanh', return_sequences=True)) m.add_layer(DenseLayer(5, 1, activation='sigmoid')) m.compile(lr=0.1) print m.predict(np.arange(6).reshape(3, 2))