最近关注了一阵Keras,感觉这个东西挺方便的,今天尝试了一下发现确实还挺方便。不但提供了常用的Layers、Normalization、Regularation、Activation等算法,甚至还包括了几个常用的数据库例如cifar-10和mnist等等。
下面的代码算是Keras的Helloworld吧!利用MLP实现的MNIST手写数字识别:
from keras.models import Sequential
from keras.layers.core import Dense, Dropout, Activation
from keras.optimizers import SGD
from keras.datasets import mnist
import numpy
model = Sequential()
model.add(Dense(784, 500, init='glorot_uniform')) # 输入层,28*28=784
model.add(Activation('tanh')) # 激活函数是tanh
model.add(Dropout(0.5)) # 采用50%的dropout
model.add(Dense(500, 500, init='glorot_uniform')) # 隐层节点500个
model.add(Activation('tanh'))
model.add(Dropout(0.5))
model.add(Dense(500, 10, init='glorot_uniform')) # 输出结果是10个类别,所以维度是10
model.add(Activation('softmax')) # 最后一层用softmax
sgd = SGD(lr=0.01, decay=1e-6, momentum=0.9, nesterov=True) # 设定学习率(lr)等参数
model.compile(loss='categorical_crossentropy', optimizer=sgd, class_mode='categorical') # 使用交叉熵作为loss函数
(X_train, y_train), (X_test, y_test) = mnist.load_data() # 使用Keras自带的mnist工具读取数据(第一次需要联网)
X_train = X_train.reshape(X_train.shape[0], X_train.shape[1] * X_train.shape[2]) # 由于mist的输入数据维度是(num, 28, 28),这里需要把后面的维度直接拼起来变成784维
X_test = X_test.reshape(X_test.shape[0], X_test.shape[1] * X_test.shape[2])
Y_train = (numpy.arange(10) == y_train[:, None]).astype(int) # 参考上一篇文章,这里需要把index转换成一个one hot的矩阵
Y_test = (numpy.arange(10) == y_test[:, None]).astype(int)
# 开始训练,这里参数比较多。batch_size就是batch_size,nb_epoch就是最多迭代的次数, shuffle就是是否把数据随机打乱之后再进行训练
# verbose是屏显模式,官方这么说的:verbose: 0 for no logging to stdout, 1 for progress bar logging, 2 for one log line per epoch.
# 就是说0是不屏显,1是显示一个进度条,2是每个epoch都显示一行数据
# show_accuracy就是显示每次迭代后的正确率
# validation_split就是拿出百分之多少用来做交叉验证
model.fit(X_train, Y_train, batch_size=200, nb_epoch=100, shuffle=True, verbose=1, show_accuracy=True, validation_split=0.3)
print 'test set'
model.evaluate(X_test, Y_test, batch_size=200, show_accuracy=True, verbose=1)
屏显输出了这么一大堆东西:
ssh://shibotian@***.***.***.***:22/usr/bin/python -u /usr/local/shared_dir/local/ipython_shibotian/shibotian/code/kreas_test1/run.py
Using gpu device 0: Tesla K40m
Train on 42000 samples, validate on 18000 samples
Epoch 0
0/42000 [==============================] - 1s - loss: 0.9894 - acc.: 0.7386 - val. loss: 0.4795 - val. acc.: 0.8807
Epoch 1
0/42000 [==============================] - 1s - loss: 0.5635 - acc.: 0.8360 - val. loss: 0.4084 - val. acc.: 0.8889
省略。。。。。
Epoch 98
0/42000 [==============================] - 1s - loss: 0.2838 - acc.: 0.9116 - val. loss: 0.1872 - val. acc.: 0.9418
Epoch 99
0/42000 [==============================] - 1s - loss: 0.2740 - acc.: 0.9163 - val. loss: 0.1842 - val. acc.: 0.9434
test set
0/10000 [==============================] - 0s - loss: 0.1712 - acc.: 0.9480
Process finished with exit code 0
P.S. verbose=1时候的进度条很可爱啊