1.导入依赖包

import tensorflow as tf

import matplotlib.pyplot as plt

2.加载数据集

(x_train, y_train), (x_test, y_test) = tf.keras.datasets.fashion_mnist.load_data()

3.对图片进行归一化

x_train, x_test = x_train / 255.0, x_test / 255.0

4.搭建模型

model = tf.keras.models.Sequential([

    tf.keras.layers.Flatten(input_shape=(28, 28)),        #Flatten：对图片进行展平    input_shape: 输入图片的尺寸

    tf.keras.layers.Dense(128, activation='relu'),         #Dense: 全连接层    128：神经元个数    activation：激活函数

    tf.keras.layers.Dropout(0.3),                                   #Dropout：随机失活神经元，防止过拟合

    tf.keras.layers.Dense(10, activation='softmax')      #10：一共有几个分类

])

5.指定优化函数、损失函数、评价指标

model.compile(optimizer='adam',                             #optimizer：优化器

              loss='sparse_categorical_crossentropy',      #loss：损失函数

              metrics=['accuracy'])                                    #评价指标

6.训练模型

history = model.fit(x_train, y_train,                             #训练数据

                    epochs=10,                                             #epochs: 训练次数

                    validation_data=(x_test, y_test),             #validation_data: 验证集

                    verbose=2)                                             #verbose：显示日志等级

Epoch 1/10
1875/1875 - 4s - loss: 0.5542 - accuracy: 0.8038 - val_loss: 0.4297 - val_accuracy: 0.8462
Epoch 2/10
1875/1875 - 4s - loss: 0.4191 - accuracy: 0.8484 - val_loss: 0.4091 - val_accuracy: 0.8519
Epoch 3/10
1875/1875 - 4s - loss: 0.3877 - accuracy: 0.8587 - val_loss: 0.3817 - val_accuracy: 0.8657
Epoch 4/10
1875/1875 - 4s - loss: 0.3673 - accuracy: 0.8663 - val_loss: 0.3745 - val_accuracy: 0.8642
Epoch 5/10
1875/1875 - 4s - loss: 0.3519 - accuracy: 0.8712 - val_loss: 0.3675 - val_accuracy: 0.8672
Epoch 6/10
1875/1875 - 3s - loss: 0.3417 - accuracy: 0.8749 - val_loss: 0.3700 - val_accuracy: 0.8674
Epoch 7/10
1875/1875 - 4s - loss: 0.3313 - accuracy: 0.8773 - val_loss: 0.3693 - val_accuracy: 0.8716
Epoch 8/10
1875/1875 - 4s - loss: 0.3259 - accuracy: 0.8785 - val_loss: 0.3525 - val_accuracy: 0.8751
Epoch 9/10
1875/1875 - 4s - loss: 0.3143 - accuracy: 0.8842 - val_loss: 0.3530 - val_accuracy: 0.8776
Epoch 10/10
1875/1875 - 4s - loss: 0.3101 - accuracy: 0.8859 - val_loss: 0.3581 - val_accuracy: 0.8751

7.显示模型准确率的变化过程

plt.plot(history.epoch, history.history.get('accuracy'), label='accuracy')                 #横坐标、纵坐标、标签

plt.plot(history.epoch, history.history.get('val_accuracy'), label='val_accuracy')

plt.legend()       #绘图

plt.show()         #显示图像

8.显示模型损失的变化过程

plt.plot(history.epoch, history.history.get('loss'), label='loss')

plt.plot(history.epoch, history.history.get('val_loss'), label='val_loss')

plt.legend()

plt.show()