DNN识别mnist手写数字

mnist数据下载地址：
链接：https://pan.baidu.com/s/1GD2hI8Wf4oUR-V2NysYorw
提取码：sg3f

导库

import numpy as np
import matplotlib.pyplot as plt
import tensorflow as tf
from tensorflow import keras
from tensorflow.keras import layers

读取mnist数据

import numpy as np
path='./mnist.npz'
f = np.load(path)
train_x, train_y = f['x_train'], f['y_train']    # 训练集
test_x, test_y = f['x_test'], f['y_test']    # 测试集
f.close()

查看数据格式

print(train_x.shape)
print(train_y.shape)
print(test_x.shape)
print(test_y.shape)

将数据以图片形式输出

plt.imshow(train_x[10000])

将数据格式改为DNN可接收的一维格式

train_x = train_x.reshape((60000,28*28),order='C')    # 将二维的图片展开为一维的数据(训练集)  
test_x = test_x.reshape((10000,28*28),order='C')    # 将二维的图片展开为一维的数据（测试集）

搭建DNN并训练

model = keras.Sequential()
model.add(layers.Dense(100,activation='relu',input_dim=28*28))
model.add(layers.Dense(10,activation='softmax'))
adam = keras.optimizers.Adam(lr=0.01)
model.compile(optimizer=adam,loss='sparse_categorical_crossentropy',metrics=['acc'])
model.fit(train_x,train_y,epochs=50,batch_size=512)

经过50轮训练后，DNN在训练集上的loss和准确率如下

DNN在测试集上的loss和准确率如下

model.evaluate(test_x,test_y)

完整的代码如下

import numpy as np
import matplotlib.pyplot as plt
import tensorflow as tf
from tensorflow import keras
from tensorflow.keras import layers

path='./mnist.npz'
f = np.load(path)
train_x, train_y = f['x_train'], f['y_train']    # 训练集
test_x, test_y = f['x_test'], f['y_test']    # 测试集
f.close()

print(train_x.shape)
print(train_y.shape)
print(test_x.shape)
print(test_y.shape)

plt.imshow(train_x[10000])

train_x = train_x.reshape((60000,28*28),order='C')    # 将二维的图片展开为一维的数据(训练集)  
test_x = test_x.reshape((10000,28*28),order='C')    # 将二维的图片展开为一维的数据（测试集）

model = keras.Sequential()
model.add(layers.Dense(100,activation='relu',input_dim=28*28))
model.add(layers.Dense(10,activation='softmax'))
model.compile(optimizer='adam',loss='sparse_categorical_crossentropy',metrics=['acc'])
model.fit(train_x,train_y,epochs=50,batch_size=512)

model.evaluate(test_x,test_y)