深度学习：Keras实现手写训练

最新推荐文章于 2025-07-04 09:30:36 发布

科学楼守夜者

最新推荐文章于 2025-07-04 09:30:36 发布

阅读量297

点赞数

CC 4.0 BY-SA版权

本文链接：https://blog.youkuaiyun.com/qq_37018566/article/details/87627809

本文介绍如何使用Keras深度学习框架，通过两层神经网络对MNIST手写数据集进行训练。详细展示了数据预处理、模型构建、训练及评估的过程。

摘要生成于 C知道，由 DeepSeek-R1 满血版支持，前往体验 >

基于Keras对MNIST手写数据集进行训练
使用两层神经网络

from keras.datasets import mnist
import keras # 深度学习框架
import keras.models # 模型
from keras.models import Sequential # 神经网络
from keras.layers import Dense, Dropout, Flatten  # 处理神经网络层
from keras.layers import Conv2D, MaxPool2D # 处理平面数据
from keras import backend as K # 处理结束


batch_size = 128 # 批量操作的大小
num_classes = 10 # 识别的结果有十类
epochs = 12
# 图片大小
img_rows, img_cols = 28, 28
(x_train, y_train), (x_test, y_test) = mnist.load_data() # 载入数据
if K.image_data_format() == "channels_first": # 格式
    x_train = x_train.reshape(x_train.shape[0], 1, img_rows, img_cols) # 形状调整
    x_test = x_test.reshape(x_test.shape[0], 1, img_rows, img_cols)
    input_shape = (1, img_rows, img_cols) # y用于训练数据形状 1*28*28
else:
    x_train = x_train.reshape(x_train.shape[0], img_rows, img_cols, 1) # 形状调整
    x_test = x_test.reshape(x_test.shape[0], img_rows, img_cols, 1)
    input_shape = (img_rows, img_cols, 1) # y用于训练数据形状 1*28*28
x_train = x_train.astype("float32")
x_test = x_test.astype("float32")
x_train /= 255 # xtrain矩阵的浮点类型， 0-255 256， 归一化
x_test /= 255
print(x_train.shape[0], "样本数量")
print(x_test.shape[0], "样本数量")
print(x_train.shape, x_test.shape)

# 结果数据，0-9，10类，分类10个输出结果
y_train = keras.utils.np_utils.to_categorical(y_train, num_classes)
y_test = keras.utils.np_utils.to_categorical(y_test, num_classes)

model = Sequential() #新建一个神经网络
model.add(Conv2D(32,
                 activation="relu",
                 input_shape=input_shape,
                 nb_row=3,
                 nb_col=3))
model.add(Conv2D(64,
                 activation="relu",
                 nb_row=3,
                 nb_col=3))
model.add(MaxPool2D(pool_size=(2, 2))) # 搜索最优结果
model.add(Dropout(0.35)) # 输出系数
model.add(Flatten()) # 展平

model.add(Dense(128,activation="relu"))
model.add(Dropout(0.5)) # 输出系数

model.add(Dense(num_classes,activation="softmax"))

model.compile(loss=keras.metrics.categorical_crossentropy,  # 训练的损失函数
              optimizer=keras.optimizers.Adadelta(), # 优化器
              metrics=["accuracy"])      # 精确
model.fit(x_train, y_train,
          batch_size=batch_size,  # 批量处理的数量
          epochs= epochs,         # 训练次数
          verbose=1,              # 行为区别
          validation_data=(x_test, y_test)) # 验证的数据
score = (model.evaluate(x_test, y_test, verbose=0))
print(score) # 识别的分数