深度学习入门-4.1 AND.py 源码分析

源代码

-----------------------------------------------------------------------------------------------------

import numpy as np

＃假设是在ubuntu平台，要找到对象的位置。一般在 /usr/local/lib/python2.7/dist-packages/numpy

from keras.models import Sequential

from keras.layers.core import Activation, Dense

from keras.optimizers import SGD

from util.callback import WeightLogger

# We are going to do a simple binary AND, so we need two inputs and a single output

X = np.zeros((4, 2))
y = np.zeros(4)

# Build the truth table for AND
X[0] = [0, 0]
y[0] = 0.0
X[1] = [0, 1]
y[1] = 0.0
X[2] = [1, 0]
y[2] = 0.0
X[3] = [1, 1]
y[3] = 1.0

model = Sequential()

# We make a single layer with one output. It takes in a 2D input.
model.add(Dense(1, input_dim=2))

# Hard sigmoid is not necessary, but it usually results in firm 1/0 output so convenient for binary
model.add(Activation('hard_sigmoid'))

# Gradient descent. We don't use the defaults because for a simple model like this the Keras defaults don't work
sgd = SGD(lr=0.1, decay=1e-6, momentum=.99)

model.compile(loss='mean_squared_error', optimizer=sgd)

# Train the model and log weights
model.fit(X, y, nb_epoch=200, batch_size=4, show_accuracy=True, verbose=1, callbacks=[WeightLogger()])

# print the truth table output for our original X
print model.predict(X)

-----------------------------------------------------------------------------------------------------------------------------

1.http://docs.scipy.org/doc/numpy-1.10.1/reference/generated/numpy.zeros.html

该网站讲述 np.zeros 的含义。每次查询都可以谷歌这个函数的目的

2.http://keras.io/getting-started/sequential-model-guide/

改网站讲述了keras.models的使用说明

3.Activation('hard_sigmoid')

hard_sigmoid激活类型

4.