初识神经网络(MNIST)

1 加载Keras中的MNIST数据集

from keras.datasets import mnist
(train_images,train_labels),(test_images,test_labels)=mnist.load_data()

train_images.shape

(60000, 28, 28)

train_labels.shape

(60000,)

train_labels

array([5, 0, 4, ..., 5, 6, 8], dtype=uint8)

2 网络架构

• 将训练数据输入到神经网络中；
• 网络学习将图像和标签关联在一起；
• 网络对test_image进行预测，验证这些预测是否与test_labels中的标签是否匹配

from keras import layers
from keras import models

networks=models.Sequential()
networks.add(layers.Dense(512,activation='relu',input_shape=(28*28,)))
networks.add(layers.Dense(10,activation='softmax'))

网络含两个Dense层，它们是密集连接（全连接），第二层是一个10路的softmax层，它返回一个由10个概率值（总和为1）组成的数组

3 编译步骤

想训练网络，我们还需要选择编译(compile)步骤的三个参数

• 损失函数
• 优化器
• 在训练和测试过程中需要监控的指标

networks.compile(optimizer='rmsprop',
                loss='categorical_crossentropy',
                metrics=['accuracy'])

4 准备图像数据

train_images=train_images.reshape((60000,28*28))
train_images=train_images.astype('float32')/255

test_images=test_images.reshape((10000,28*28))
test_images=test_images.astype('float32')/255

5 准备标签

from keras.utils import to_categorical

train_labels=to_categorical(train_labels)
test_labels=to_categorical(test_labels)

networks.fit(train_images,train_labels,epochs=5,batch_size=128)

Epoch 1/5
60000/60000 [==============================] - 5s 89us/step - loss: 0.2607 - accuracy: 0.9248
Epoch 2/5
60000/60000 [==============================] - 5s 82us/step - loss: 0.1047 - accuracy: 0.9693
Epoch 3/5
60000/60000 [==============================] - 5s 82us/step - loss: 0.0686 - accuracy: 0.9792
Epoch 4/5
60000/60000 [==============================] - 6s 98us/step - loss: 0.0512 - accuracy: 0.9843
Epoch 5/5
60000/60000 [==============================] - 5s 91us/step - loss: 0.0380 - accuracy: 0.9887





<keras.callbacks.callbacks.History at 0x177b9fcb748>