线性代数与数据学习:MIT教授Gilbert Strang帮你打下坚实的数学基础

机器之心编辑,作者:思源、刘晓坤。

MIT 教授 Gilbert Strang 最新书籍《线性代数与数据学习》(Linear Algebra and Learning from Data)将在 1 月中旬发行。这一本书为机器学习提供了很多数学基础,它同时也提供了深度学习一些基本概念。可以说借助这本书,我们能从数学的角度来理解流行的模型。

书籍主页:math.mit.edu/~gs/learnin…

这本书的目的是解释数据科学和机器学习所依赖的数学:线性代数、最优化、概率论和统计学。因为在机器学习中,学习函数中的权重会以矩阵形式表示,这些权重通过随机梯度下降优化,而「随机」一词提示训练收敛是概率性的。此外,概率论中的大数定律被扩展到了大函数定律:如果架构设计良好并且参数计算良好,则有很高的概率能成功收敛。

请注意这不是一本关于计算或编码或软件的书。已经有很多书籍对这些方面做了很好的介绍,比如《Hands-On Machine Learning》;还有很多 TensorFlow、Keras、MathWorks 和 Caffe 等的在线资源,也能提供很多帮助。

线性代数有众多美妙的矩阵变体:对称矩阵、正交矩阵、三角矩阵、Banded 矩阵、转置矩阵和正定矩阵等等。在 Gilbert 的教学经验中,他认为正定对称矩阵 S 是非常美妙的东西。它们有正的特征值λ和正交的特征向量 q,它们的线性组合可以将秩为 1 的简单映射 qq^T 与对应特征值重构为正定矩阵 S,即:

如果 λ_1>=λ_2>=...,那么上式特征值λ_1 以及对应的特征向量组成的第一个分量就是 S 最具信息的部分。对于一个简单的协方差矩阵,这一部分就对应着对大的方差,这也是降维算法 PCA 最核心的思想。

此外,在书籍主页中,作者还提供了试读的样章,包括深度学习、书籍前言、目录、矩阵初等变换、矩阵乘法和其它一些从矩阵看卷积网络等新知识。作者表明书籍主页会持续更新,包括印刷计划和全本开放阅读等。

William Gilbert Strang

William Gilbert Strang,美国数学家,在有限元理论、变分法、小波分析和线性代数等方面皆有研究贡献。他对数学教育做出了许多贡献,包括出版七本数学教科书和专著。斯特朗现任麻省理工学院数学系 MathWorks 讲座教授。主要讲授课程为线性代数入门(Introduction to Linear Algebra,18.06)和计算科学与工程(Computational Science and Engineering,18.085),这些课程都可在麻省理工学院开放式课程中免费学习。

以下是这本书的目录:



Machine learning allows computational systems to adaptively improve their performance with experience accumulated from the observed data. Its techniques are widely applied in engineering, science, finance, and commerce. This book is designed for a short course on machine learning. It is a short course, not a hurried course. From over a decade of teaching this material, we have distilled what we believe to be the core topics that every student of the subject should know. We chose the title `learning from data' that faithfully describes what the subject is about, and made it a point to cover the topics in a story-like fashion. Our hope is that the reader can learn all the fundamentals of the subject by reading the book cover to cover. ---- Learning from data has distinct theoretical and practical tracks. In this book, we balance the theoretical and the practical, the mathematical and the heuristic. Our criterion for inclusion is relevance. Theory that establishes the conceptual framework for learning is included, and so are heuristics that impact the performance of real learning systems. ---- Learning from data is a very dynamic field. Some of the hot techniques and theories at times become just fads, and others gain traction and become part of the field. What we have emphasized in this book are the necessary fundamentals that give any student of learning from data a solid foundation, and enable him or her to venture out and explore further techniques and theories, or perhaps to contribute their own. ---- The authors are professors at California Institute of Technology (Caltech), Rensselaer Polytechnic Institute (RPI), and National Taiwan University (NTU), where this book is the main text for their popular courses on machine learning. The authors also consult extensively with financial and commercial companies on machine learning applications, and have led winning teams in machine learning competitions.
Preface I wrote this book to help machine learning practitioners, like you, get on top of linear algebra, fast. Linear Algebra Is Important in Machine Learning There is no doubt that linear algebra is important in machine learning. Linear algebra is the mathematics of data. It’s all vectors and matrices of numbers. Modern statistics is described using the notation of linear algebra and modern statistical methods harness the tools of linear algebra. Modern machine learning methods are described the same way, using the notations and tools drawn directly from linear algebra. Even some classical methods used in the field, such as linear regression via linear least squares and singular-value decomposition, are linear algebra methods, and other methods, such as principal component analysis, were born from the marriage of linear algebra and statistics. To read and understand machine learning, you must be able to read and understand linear algebra. Practitioners Study Linear Algebra Too Early If you ask how to get started in machine learning, you will very likely be told to start with linear algebra. We know that knowledge of linear algebra is critically important, but it does not have to be the place to start. Learning linear algebra first, then calculus, probability, statistics, and eventually machine learning theory is a long and slow bottom-up path. A better fit for developers is to start with systematic procedures that get results, and work back to the deeper understanding of theory, using working results as a context. I call this the top-down or results-first approach to machine learning, and linear algebra is not the first step, but perhaps the second or third. Practitioners Study Too Much Linear Algebra When practitioners do circle back to study linear algebra, they learn far more of the field than is required for or relevant to machine learning. Linear algebra is a large field of study that has tendrils into engineering, physics and quantum physics. There are also theorems and derivations for nearly everything, most of which will not help you get better skill from or a deeper understanding of your machine learning model. Only a specific subset of linear algebra is required, though you can always go deeper once you have the basics.
评论
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值