本书全面介绍TensorFlow 2.x 框架及其在深度学习中的应用，内容包括TensorFlow 简介、Python 语 言基础、环境搭建与入门、TensorBoard 可视化、多层感知机实现、卷积神经网络实现、循环神经网络实 现、强化学习、迁移学习、生成对抗网络和GPU 并行计算等。

第1 章 TensorFlow 简介 ·············································································.1
1.1 人工智能的编程框架 ................................................................................................. 1
1.1.1 人工智能的发展 ............................................................................................. 1
1.1.2 人工智能、机器学习和深度学习之间的关系 ............................................. 2
1.2 TensorFlow 与人工智能 ............................................................................................ 3
1.3 TensorFlow 数据模型 ................................................................................................ 4
1.4 TensorFlow 计算模型和运行模型 ............................................................................ 5
1.5 实验：矩阵运算 ......................................................................................................... 9
1.5.1 实验目的 ......................................................................................................... 9
1.5.2 实验要求 ......................................................................................................... 9
1.5.3 实验原理 ......................................................................................................... 9
1.5.4 实验步骤 ....................................................................................................... 10
习题 .................................................................................................................................... 10
第2 章 Python 语言基础 ············································································.11
2.1 Python 语言 ............................................................................................................... 11
2.1.1 Python 语言的发展 ....................................................................................... 11
2.1.2 Python 安装 ................................................................................................... 12
2.2 基础语法 ................................................................................................................... 13
2.2.1 基础知识 ....................................................................................................... 13
2.2.2 基本程序编写 ............................................................................................... 15
2.2.3 条件语句 ....................................................................................................... 16
2.2.4 循环语句 ....................................................................................................... 17
2.3 数据结构 ................................................................................................................... 18
2.4 面向对象特性 ........................................................................................................... 21
2.4.1 类和对象 ....................................................................................................... 21
2.4.2 类的定义 ....................................................................................................... 22
2.4.3 根据类创建对象 ........................................................................................... 22
2.4.4 构造方法与析构方法 ................................................................................... 23
2.5 其他高级特性 ........................................................................................................... 24
2.5.1 函数高级特性 ............................................................................................... 24
2.5.2 闭包 ............................................................................................................... 25
2.6 实验：Python 基本语法的实现 ............................................................................... 26
2.6.1 实验目的 ....................................................................................................... 26
2.6.2 实验要求 ....................................................................................................... 26
2.6.3 实验题目 ....................................................................................................... 26
2.6.4 实验步骤 ....................................................................................................... 27
习题 .................................................................................................................................... 28
第3 章 环境搭建与入门 ·············································································.30
3.1 开发平台简介 ........................................................................................................... 30
3.2 开发环境部署 ........................................................................................................... 30
3.2.1 安装Anaconda .............................................................................................. 30
3.2.2 安装TensorFlow ........................................................................................... 32
3.2.3 PyCharm 下载与安装 ................................................................................... 32
3.3 一个简单的实例 ....................................................................................................... 34
习题 .................................................................................................................................... 36
第4 章 TensorBoard 可视化 ········································································.37
4.1 什么是TensorBoard.................................................................................................. 37
4.2 基本流程与结构 ....................................................................................................... 37
4.3 图表的可视化 ........................................................................................................... 39
4.3.1 计算图和会话 ............................................................................................... 39
4.3.2 可视化过程 ................................................................................................... 40
4.4 监控指标的可视化 ................................................................................................... 41
4.4.1 Scalar ............................................................................................................. 41
4.4.2 Images ........................................................................................................... 41
4.4.3 Histogram ...................................................................................................... 41
4.4.4 Merge_all....................................................................................................... 42
4.5 学习过程的可视化 ................................................................................................... 42
4.5.1 数据序列化 ................................................................................................... 43
4.5.2 启动TensorBoard ......................................................................................... 43
4.6 实验：TensorBoard 可视化实现 .............................................................................. 44
4.6.1 实验目的 ....................................................................................................... 44
4.6.2 实验要求 ....................................................................................................... 44
4.6.3 实验原理 ....................................................................................................... 45
4.6.4 实验步骤 ....................................................................................................... 45
习题 .................................................................................................................................... 49
第5 章 多层感知机实现 ·············································································.50
5.1 感知机 ....................................................................................................................... 50
5.1.1 感知机的定义 ............................................................................................... 50
5.1.2 感知机的神经元模型 ................................................................................... 51
5.1.3 感知机的学习算法 ....................................................................................... 51
5.1.4 感知机的性质 ............................................................................................... 52
5.2 多层感知机与前向传播 ........................................................................................... 53
5.2.1 多层感知机基本结构 ................................................................................... 53
5.2.2 多层感知机的特点 ....................................................................................... 54
5.3 前向传播 ................................................................................................................... 55
5.3.1 前向传播的计算过程 ................................................................................... 55
5.3.2 前向传播算法 ............................................................................................... 57
5.4 梯度下降 ................................................................................................................... 57
5.4.1 梯度 ............................................................................................................... 57
5.4.2 梯度下降的直观解释 ................................................................................... 58
5.4.3 梯度下降法的相关概念 ............................................................................... 58
5.4.4 梯度下降法的数学描述 ............................................................................... 59
5.4.5 梯度下降法的算法调优 ............................................................................... 60
5.4.6 常见的梯度下降法 ....................................................................................... 60
5.5 反向传播 ................................................................................................................... 61
5.5.1 反向传播算法要解决的问题 ....................................................................... 61
5.5.2 反向传播算法的基本思路 ........................................................................... 61
5.5.3 反向传播算法的流程 ................................................................................... 63
5.6 数据集 ....................................................................................................................... 64
5.6.1 训练集、测试集和验证集 ........................................................................... 64
5.6.2 MNIST 数据集 ............................................................................................. 64
5.7 多层感知机的实现 ................................................................................................... 66
5.7.1 NumPy 多层感知机的实现 .......................................................................... 66
5.7.2 TensorFlow 多层感知机的实现 ................................................................... 69
5.8 实验：基于Keras 多层感知机的MNIST 手写数字识别 ...................................... 72
5.8.1 Keras 简介 ..................................................................................................... 72
5.8.2 实验目的 ....................................................................................................... 73
5.8.3 实验要求 ....................................................................................................... 73
5.8.4 实验步骤 ....................................................................................................... 73
习题 .................................................................................................................................... 77
第6 章 卷积神经网络实现 ··········································································.78
6.1 CNN 基本原理 .......................................................................................................... 78
6.2 CNN 的卷积操作 ...................................................................................................... 80
6.3 CNN 的池化操作 ...................................................................................................... 82
6.4 使用简单的CNN 实现手写字符识别 ..................................................................... 84
6.5 AlexNet ..................................................................................................................... 85
6.6 实验：基于VGG16 模型的图像分类实现 ............................................................. 87
6.6.1 实验目的 ....................................................................................................... 87
6.6.2 实验要求 ....................................................................................................... 87
6.6.3 实验原理 ....................................................................................................... 88
6.6.4 实验步骤 ....................................................................................................... 88
习题 .................................................................................................................................... 93
第7 章 循环神经网络实现 ··········································································.94
7.1 RNN 简介 .................................................................................................................. 94
7.1.1 为什么使用RNN.......................................................................................... 94
7.1.2 RNN 的网络结构及原理 .............................................................................. 96
7.1.3 RNN 的实现 ................................................................................................. 99
7.2 长短时记忆网络 ..................................................................................................... 100
7.2.1 长期依赖问题 ............................................................................................. 100
7.2.2 长短时记忆网络 ......................................................................................... 101
7.2.3 LSTM 的实现 ............................................................................................. 105
7.3 双向RNN ................................................................................................................ 106
7.3.1 双向RNN 的结构及原理 ........................................................................... 106
7.3.2 双向RNN 的实现....................................................................................... 107
7.4 深层RNN ................................................................................................................ 108
7.5 实验：基于LSTM 的股票预测 ............................................................................. 110
7.5.1 实验目的 ..................................................................................................... 110
7.5.2 实验要求 ..................................................................................................... 110
7.5.3 实验原理 ..................................................................................................... 111
7.5.4 实验步骤 ..................................................................................................... 111
习题 .................................................................................................................................. 114
第8 章 强化学习 ····················································································.115
8.1 强化学习原理 ......................................................................................................... 115
8.2 马尔可夫决策过程实现 ......................................................................................... 117
8.2.1 马尔可夫决策过程 ..................................................................................... 117
8.2.2 马尔可夫决策过程的形式化 ..................................................................... 118
8.3 基于价值的强化学习方法 ..................................................................................... 120
8.3.1 基于价值的方法中的策略优化 ................................................................. 120
8.3.2 基于价值的方法中的策略评估 ................................................................. 120
8.3.3 Q-Learning .................................................................................................. 122
8.4 Gym 的简单使用 .................................................................................................... 123
8.5 实验：基于强化学习的小车爬山游戏 ................................................................. 125
8.5.1 实验目的 ..................................................................................................... 125
8.5.2 实验要求 ..................................................................................................... 125
8.5.3 实验原理 ..................................................................................................... 125
8.5.4 实验步骤 ..................................................................................................... 127
习题 .................................................................................................................................. 130
第9 章 迁移学习 ····················································································.131
9.1 迁移学习原理 ......................................................................................................... 131
9.1.1 什么是迁移学习 ......................................................................................... 131
9.1.2 迁移学习的基本概念 ................................................................................. 131
9.1.3 迁移学习的基本方法 ................................................................................. 133
9.2 基于模型的迁移学习方法实现 ............................................................................. 134
9.2.1 导入已有的预训练模型 ............................................................................. 134
9.2.2 模型的复用 ................................................................................................. 134
9.2.3 基于新模型的预测 ..................................................................................... 135
9.3 基于VGG-19 的迁移学习实现 ............................................................................. 135
9.3.1 VGG-19 的原理 .......................................................................................... 135
9.3.2 基于VGG-19 的迁移学习的原理及实现 ................................................. 136
9.4 实验：基于Inception V3 的迁移学习 .................................................................. 138
9.4.1 实验目的 ..................................................................................................... 138
9.4.2 实验要求 ..................................................................................................... 138
9.4.3 实验原理 ..................................................................................................... 139
9.4.4 实验步骤 ..................................................................................................... 140
习题 .................................................................................................................................. 143
第10 章 生成对抗网络 ·············································································.144
10.1 GAN 概述 ............................................................................................................. 144
10.2 GAN 的目标函数 ................................................................................................. 144
10.3 GAN 的实现 ......................................................................................................... 145
10.4 深度卷积生成对抗网络 ....................................................................................... 149
10.4.1 DCGAN 结构图 ........................................................................................ 150
10.4.2 DCGAN 的实现 ........................................................................................ 150
10.5 GAN 的衍生模型 ................................................................................................. 153
10.5.1 基于网络结构的衍生模型 ....................................................................... 154
10.5.2 基于优化方法的衍生模型 ....................................................................... 155
习题 .................................................................................................................................. 156
第11 章 GPU 并行计算 ············································································.157
11.1 并行计算技术 ....................................................................................................... 157
11.1.1 单机并行计算 ........................................................................................... 157
11.1.2 分布式并行计算 ....................................................................................... 158
11.1.3 GPU 并行计算技术 .................................................................................. 159
11.1.4 TensorFlow 与GPU .................................................................................. 160
11.2 TensorFlow 加速方法 ........................................................................................... 163
11.3 单GPU 并行加速的实现 ..................................................................................... 170
11.4 多GPU 并行加速的实现 ..................................................................................... 173
11.5 实验：基于GPU 的矩阵乘法 ............................................................................. 175
11.5.1 安装GPU 版本的TensorFlow ................................................................. 175
11.5.2 一个GPU 程序 ......................................................................................... 176
11.5.3 使用GPU 完成矩阵乘法 ......................................................................... 176
习题 .................................................................................................................................. 177