AI (Artificial Intelligence, AI) is a broad field covering a number of subordinate disciplines, this series of books covering a particular topic among some, and this book is the first volume of the book series. The next few sections will present a series of books and introduced this Volume.
Series book about
This series of books will introduce the reader to various hot topics in the field of artificial intelligence. Since AI is a large and complicated field, and which covers growing, any book could only focus on specific areas, so this book is also intended to be a tremendous amount of Artificial Intelligence tutorial.
This series of books on a mathematical easy way to understand the concepts taught in artificial intelligence, which is the English title of this series of books in the "for Human" meaning. In addition:
- This series of books assume the reader proficient in at least one programming language;
- This series of books assume the reader has a basic understanding of algebra University;
- This series of books will use calculus, linear algebra, differential equations and statistics related concepts and formulas;
- But in interpreting the content of the third point, this series of books does not assume that the reader of the above very skilled;
- All concepts are not just mathematical formulas, but also with programming examples and the pseudo-code.
Audience This series of books is a programmer proficient in at least one programming language, and the examples in the book have been rewritten to form a variety of programming languages.
Volume 1 artificial intelligence algorithms based algorithm in pseudo-code gives only, and specific examples of the code places Java, C #, R, C / C ++ and Python language form, in addition to the community maintenance Scala supported languages. Community members are trying to convert the sample code to other programming languages more, maybe when you get the book, also has sample code version of your favorite programming language. Access book GitHub open source warehouse to get more information, we also encourage community collaboration to help us complete rewriting the code and porting. If you want to join the collaboration, we would be grateful. More information can be found in the process of this book Appendix A.
Series books publishing program
This series of books written plan:
- Volume 0: Getting Started AI mathematics;
- Volume 1: Basic algorithm;
- Volume 2: Nature heuristic algorithm;
- Volume 3: neural networks;
- Volume 4: support vector machine;
- Volume 5: probability learning.
Volume 1 to Volume 5 will be followed by the publication; and 0 will roll as "ahead of a planned prequel" to complete this series of books published near the end of the occasion. Volume 1 to Volume 5 volume will explain the necessary mathematical concepts, Volume 0 will focus on the review of these concepts, and some expansion on this basis.
0 volumes in this series may be a beginning of the book, may be summarized as a series of books; reading sequential volume is preferably 1 before subsequent volumes; 2 volume parts of the reader in understanding the contents of the volume 3 there is helpful . Figure 1 shows the logical reading order of our recommendations.
Volume 1 algorithm based artificial intelligence algorithms
Jeffrey Heaton (Jeffery Heaton) with, over Lear translation
- AI algorithms introductory tutorial books, everyone can read the book Artificial Intelligence
- Full-color printing, easy-to-understand examples to explain the basis of artificial intelligence algorithms
- The sample code in multiple languages, a wealth of online resources, and expand to facilitate hands-on practical learning
Each volume in this series of books can be read independently, but also as a whole series of books to read. But note that, Volume 1 lists all the basic algorithm used for subsequent volumes, and these are both based algorithm itself, without losing practicality.
1 Volume flow reading head
The basic algorithm introduced
Yu Jian rise, will re-foundation. This book will teach methods such as dimensions, distance measure algorithm, clustering algorithm, error calculation, hill-climbing algorithm, linear regression and discrete learning such as artificial intelligence algorithms. These correspond to the data processing algorithms and identify a particular pattern, but also as a logical Amazon (Amazon) network and fly (Netflix) such sites, various recommendation system behind.
These algorithms are not only the basis for the algorithm described in subsequent volumes, which itself is of great use. In this book, these algorithms are attached examples to explain the highly workable numerical calculation.
Volume 1 content structure
Chapter 1 "AI Getting Started", introduces the basic concepts of the book or series of books other volumes will be used in part related to artificial intelligence. Most artificial intelligence algorithms that accepts an input array to produce an output array - can solve the problem of artificial intelligence are usually classified as such models. In the internal algorithm models also we need to have additional array to store long-term memory. Training algorithm is actually a process to produce the expected output corresponding to a given input by adjusting the value of long-term memory.
第2章“数据归一化”，描述了大多数人工智能算法对原始数据的预处理流程。数据需要以一个输入数组的形式传递给算法，但实践中获取到的数据并不一定都是数值型的，也有一些是类别信息，比如颜色、形状、性别、物种抑或其他一些非数值型的描述性特征。此外，就算是现成的数值型数据，也必须在一定范围内归一化，并且通常是归一化到 (-1, 1) 区间。
第3章“距离度量”，展示了我们比较数据的方法，说起来这种比较方法其实跟在地图上标识出两点间的距离十分相像。人工智能通常以数值数组的形式处理数据，包括输入数据、输出数据、长期记忆、短期记忆和其他很多数据都是如此，这些数组很多时候也被称作“向量”。我们可以像计算两点间距离一样，计算出两个数据之间的差异（二维和三维的点可以分别看作长度为二和三的向量）。当然，在人工智能领域，我们经常要处理的是更高维空间中的数据。
第4章“随机数生成”，讲解了人工智能算法中随机数的生成和使用。本章由关于均匀随机数和正态随机数的讨论切入——出现这种不同的根源在于有的时候算法要求随机数具有等可能性，而有的时候又需要它们服从某种既定的分布。此外本章还讨论了生成随机数的方法。
第5章“K均值聚类算法”，详述了将数据按相似度分类的方法。K均值算法本身可以用来将数据按共性分组，同时也可以被用于组成更复杂的算法——比如遗传算法就利用K均值算法对种群按特征归类，各路网商也利用聚类算法划分顾客，依照同类型顾客的消费习惯调整销售策略。
第6章“误差计算”，演示了评估人工智能算法效果的方法。误差计算的过程由一个用以评估算法最终效果的评分函数执行，其结果决定了算法的效果。一类常用的评分函数只需要给定输入向量和预期输出向量，也就是所谓的“训练数据”，算法的效果则是由实际输出与预期输出间的差异决定的。
第7章“迈向机器学习”，概述了可以从数据中学习特征来优化结果的简单机器学习算法。大多数人工智能算法是用权值向量将输入向量转化为期望的输出向量，这些权值向量构成了算法的长期记忆，“训练”就是一个调整长期记忆以产生预期输出的过程。本章会演示几个具有学习能力的简单模型的构建方法，也会介绍一些简单但却行之有效的训练算法，能够调整这种长期记忆（权重向量）并优化输出结果，简单随机漫步和爬山算法正是其中之二。
第8章“优化训练”，在前面章节的基础上进行了一定的拓展，介绍了像模拟退火算法和Nelder-Mead法[2]这样用来快速优化人工智能模型权重的算法。本章还说明了如何通过一定的调整，将这些优化算法应用于之前提到过的部分模型。
第9章“离散优化”，解释了如何优化非数值型的类别型数据。并非所有优化问题都是数值型的，还有离散型和类别型问题，比如背包问题和旅行商问题。本章将说明模拟退火算法可以用于处理这两个问题，并且该算法既适用于连续的数值型问题，也适用于离散的类别型问题。
第10章“线性回归”，讲解了如何用线性和非线性方程来学习趋势并做出预测。本章将介绍简单线性回归，并演示如何用它来拟合数据为线性模型。此外还将介绍可以拟合非线性数据的广义线性模型（General Linear Model，GLM）。
目录结构：
第 1 章 AI 入门 1
1．1 与人类大脑的联系 2
1．2 对问题建模 6
1．3 对输入/ 输出建模 11
1．4 理解训练过程 21
1．5 本章小结 23
第 2 章 数据归一化 25
2．1 计量尺度 25
2．2 观测值归一化 29
2．3 其他归一化方法 38
2．4 本章小结 45
第3 章 距离度量 47
3．1 理解向量 47
3．2 计算向量距离 49
3．3 光学字符识别 54
3．4 本章小结 57
第4 章 随机数生成 59
4．1 伪随机数生成算法的概念 60
4．2 随机数分布类型 61
4．3 轮盘模拟法 64
4．4 伪随机数生成算法 65
4．5 用蒙特卡洛方法估算PI 值 72
4．6 本章小结 74
第5 章 K 均值聚类算法 75
5．1 理解训练集 77
5．2 理解K 均值算法 80
5．3 K 均值算法的初始化 84
5．4 本章小结 90
第6 章 误差计算 91
6．1 方差和误差 92
6．2 均方根误差 93
6．3 均方误差 93
6．4 误差计算方法的比较 94
6．5 本章小结 96
第7 章 迈向机器学习 97
7．1 多项式系数 99
7．2 训练入门 101
7．3 径向基函数网络 103
7．4 本章小结 115
第8 章 优化训练 117
8．1 爬山算法 117
8．2 模拟退火算法 121
8．3 Nelder-Mead 算法 128
8．4 Nelder-Mead 算法的终止条件 133
8．5 本章小结 134
第9 章 离散优化 135
9．1 旅行商问题 135
9．2 环形旅行商问题 138
9．3 背包问题 139
9．4 本章小结 143
第 10 章 线性回归 144
10．1 线性回归 144
10．2 广义线性模型 152
10．3 本章小结 155
附录A 示例代码使用说明 157
A．1 “读懂人工智能”系列书简介 157
A．2 保持更新 157
A．3 获取示例代码 158
A．4 示例代码的内容 159
A．5 如何为项目做贡献 163
参考资料 164
