《Python 深度学习》刷书笔记 Chapter 2 构建一个简单的神经网络

本文链接：https://blog.csdn.net/LeungSr/article/details/120583917

本文介绍了使用Keras库构建神经网络处理MNIST手写数字识别任务的过程。从加载数据集到预处理图像，再到构建模型、编译模型、训练模型以及评估模型，详细展示了每一步的操作和结果。最后，讨论了张量的切片操作。

摘要由CSDN通过智能技术生成

文章目录

2.2 神经网络的表示
写在最后

2.2 神经网络的表示

代码清单 2-1

# 解决编译器报错 using tenserflow backend的问题
import os
os.environ['KERAS_BACKEND']='tensorflow'

from keras.datasets import mnist

(train_images,train_labels),(test_images,test_labels) = mnist.load_data()

# 训练集大小
train_images.shape
len(train_labels)

# 训练集标签
train_labels

# 测试集
test_images.shape
len(test_labels)

# 测试集标签
test_labels

array([7, 2, 1, ..., 4, 5, 6], dtype=uint8)

代码清单 2-2, 2-3

from keras import models
from keras import layers
import numpy as np

# 神经网络框架
network = models.Sequential()
network.add(layers.Dense(512,activation='relu',input_shape=(28*28,)))
network.add(layers.Dense(10,activation='softmax'))

# 编译步骤
network.compile(optimizer = 'rmsprop', loss = 'categorical_crossentropy', metrics=['accuracy']);

代码清单 2-4

# 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

代码清单 2-5-1

# 准备数据标签

# from keras.utils import to_categoriacl
from tensorflow.keras.utils import to_categorical

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

# 训练网络
# loss:训练数据的损失 acc：网络在训练数据上的精度
network.fit(train_images , train_labels , epochs = 5 , batch_size = 128)

Epoch 1/5
60000/60000 [==============================] - 3s 43us/step - loss: 0.0283 - accuracy: 0.9917
Epoch 2/5
60000/60000 [==============================] - 3s 44us/step - loss: 0.0216 - accuracy: 0.9937
Epoch 3/5
60000/60000 [==============================] - 3s 42us/step - loss: 0.0173 - accuracy: 0.9951
Epoch 4/5
60000/60000 [==============================] - 3s 44us/step - loss: 0.0127 - accuracy: 0.9965
Epoch 5/5
60000/60000 [==============================] - 3s 48us/step - loss: 0.0099 - accuracy: 0.9972

代码清单 2-5-2

test_loss, test_acc = network.evaluate( test_images , test_labels)
print('test_acc:',test_acc)

10000/10000 [==============================] - 0s 28us/step
test_acc: 0.9797000288963318

代码清单 2-6

观察MNIST数据集，MNIST数据集通常用于手写识别普通的神经网络，我们结合本小节提到的张量知识来观察该数据集

# 读入数据集
from keras.datasets import mnist

(train_images, train_labels), (test_images, test_labels) = mnist.load_data()

# 显示张量mnist个数
print(train_images.ndim)

# 显示张量的形状
print(train_images.shape)

# 显示张量的数据类型
print(train_images.dtype)

3
(60000, 28, 28)
uint8

接着，我们观察MNIST数据集里面各个样本的样子
我们以第四个数字为例，显示第四个数字

digit  = train_images[4]

import matplotlib.pyplot as plt
plt.imshow(digit,cmap=plt.cm.binary)
plt.show()

在这里插入图片描述

张量切片

通过冒号，选择特定的张良新元素叫做张量切片

# 对于一个三维的张量，有三种不同的写法

# 写法一
my_slice = train_images[10:100]
print(my_slice.shape)

# 写法二
my_slice2 = train_images[10:100, :, :]
print(my_slice2.shape)
    
# 写法三
my_slice3 = train_images[10:100, 0:28, 0:28]
print(my_slice3.shape)