如何训练一个神经网络

import os,shutil
import tensorflow
from tensorflow.keras import models
from tensorflow.keras import layers
from tensorflow.keras import optimizers
from tensorflow.keras.preprocessing.image import ImageDataGenerator
import matplotlib.pyplot as plt

# 数据集
train_dir = './任务1/train'  # 训练集目录
test_dir = './任务1/test'  # 测试集目录
print('train_0_images:', len(os.listdir(os.path.join(train_dir, '0'))))
print('train_1_images:', len(os.listdir(os.path.join(train_dir, '1'))))
print('train_2_images:', len(os.listdir(os.path.join(train_dir, '2'))))
print('test_0_images:', len(os.listdir(os.path.join(test_dir, '0'))))
print('test_1_images:', len(os.listdir(os.path.join(test_dir, '1'))))
print('test_2_images:', len(os.listdir(os.path.join(test_dir, '2'))))

model = models.Sequential()
model.add(layers.Conv2D(32,(3,3),activation='relu',padding='same',input_shape=((28,28,1))))
model.add(layers.MaxPool2D(2,2))
# model.add(layers.Conv2D(64,(3,3),activation='relu'))
# model.add(layers.MaxPool2D(2,2))
# model.add(layers.Conv2D(128,(3,3),activation='relu'))
# model.add(layers.MaxPool2D(2,2))
# # model.add(layers.Conv2D(128,(3,3),activation='relu'))
# # model.add(layers.MaxPool2D(2,2))
model.add(layers.Flatten())
# model.add(layers.Dense(8,activation = 'relu'))
model.add(layers.Dense(3,activation="softmax"))
model.summary()

# 选择优化器
model.compile(loss='categorical_crossentropy',#binary_crossentropy二分类
              optimizer=optimizers.RMSprop(lr=1e-4),
              metrics=['acc'])

#数据预处理
#这里不理解可以看一下博客里面的相关的一篇文章
train_datagen = ImageDataGenerator(rescale=1./255)
test_datagen = ImageDataGenerator(rescale=1./255)
# 它生成了 150×150 的 RGB 图像
#［形状为 (20, 150, 150, 3)］
#与二进制标签［形状为 (20,)］组成的批量
train_generator = train_datagen.flow_from_directory(
    train_dir,
    target_size=(28,28),
    batch_size=20,
    color_mode='grayscale',
    class_mode='categorical'#这个地方二分类是binary，多分类是categorical
)
validation_generator = test_datagen.flow_from_directory(
    test_dir,
    target_size=(28,28),#target_size=(28,28),这里生成的图像是（28，28，3），但实际上我们的图像只有1通道
    color_mode='grayscale',
    batch_size=20,
    class_mode='categorical'
)

# 利用批量生成器拟合模型
#  steps_per_epoch 参数的作用：从生成器中抽取 
# steps_per_epoch 个批量后拟合过程
# 每个批量包含 20 个样本，所以读取完所有 2000 个样本需要 100个批量
# validation_steps:需要从验证生成器中抽取多少个批次用于评估
history = model.fit_generator(
      train_generator,
      steps_per_epoch=100,
      epochs=20,
      validation_data=validation_generator,
      validation_steps=20)

#保存模型
model.save('./任务1/small_data_1.h5')

# 绘制训练过程中的损失曲线和精度曲线
acc = history.history['acc']
val_acc = history.history['val_acc']
loss = history.history['loss']
val_loss = history.history['val_loss']

epochs = range(1,len(acc) +1)

plt.plot(epochs,acc,'bo',label='Training acc')
plt.plot(epochs,val_acc,'b',label='validation acc')
plt.title('train and validation acc')
plt.legend()

plt.figure()

plt.plot(epochs,loss,'bo',label='Training loss')
plt.plot(epochs,val_loss,'b',label='val_loss acc')
plt.title('train and validation loss')
plt.legend()

plt.show()