**Название**: Machine Learning with Python: Theory and Implementation

**Автор**: Amin Zollanvari

**Издательство**: Springer

**Год**: 2023

**Страниц**: 457

**Язык**: английский

**Формат**: pdf (true)

**Размер**: 39.1 MB

This book is meant as a textbook for undergraduate and graduate students who are willing to understand essential elements of Machine Learning from both a theoretical and a practical perspective. The choice of the topics in the book is made based on one criterion: whether the practical utility of a certain method justifies its theoretical elaboration for students with a typical mathematical background in engineering and other quantitative fields. As a result, not only does the book contain practically useful techniques, it also presents them in a mathematical language that is accessible to both graduate and advanced undergraduate students.

The primary goal of this book is to provide, as far as possible, a concise systematic introduction to practical and mathematical aspects of Machine Learning algorithms. The book has arisen from about eight years of my teaching experience in both Machine Learning and programming, as well as over 15 years of research in fundamental and practical aspects of Machine Learning. There were three sides to my main motivation for writing this book:

• Software skills: During the past several years of teaching Machine Learning as both undergraduate and graduate level courses, many students with different mathematical, statistical, and programming backgrounds have attended my lectures. This experience made me realize that many senior year undergraduate or postgraduate students in STEM (Science, Technology, Engineering, and Mathematics) often have the necessary mathematical-statistical background to grasp the basic principles of many popular techniques in Machine Learning. That being said, a major factor preventing students from deploying state-of-the-art existing software for machine learning is often the lack of programming skills in a suitable programming language. In the light of this experience, along with the current dominant role of Python in Machine Learning, I often ended up teaching for a few weeks the basic principles of Python programming to the extent that is required for using existing software. Despite many excellent practical books on machine learning with Python applications, the knowledge of Python programming is often stated as the prerequisite. This led me to make this book as self-contained as possible by devoting two chapters to fundamental Python concepts and libraries that are often required for machine learning. At the same time, throughout the book and in order to keep a balance between theory and practice, the theoretical introduction to each method is complemented by its Python implementation either in Scikit-Learn or Keras.

• Balanced treatment of various supervised learning tasks: In my experience, and insofar as supervised learning is concerned, treatment of various tasks such as binary classification, multiclass classification, and regression is rather imbalanced in the literature; that is to say, methods are often presented for binary classification with the multiclass classification being left implicit; or even though in many cases the regressor within a class of models can be easily obtained from the classifier within the same family, the focus is primarily on the classification side. Inconsistency in treating various tasks often prevents learners from connecting the dots and seeing a broader picture. In this book, I have attempted to strike a balance between the aforementioned tasks and present them whenever they are available within a class of methods.

• Focus on “core” techniques and concepts: Given the myriad of concepts and techniques used in machine learning, we should naturally be selective about the choice of topics to cover. In this regard, in each category of common practices in Machine Learning (e.g., model construction, model evaluation, etc.), I included some of the most fundamental methods that: 1) have strong track record of success in many applications; and 2) are presentable to students with an “average” STEM background—I would refer to these as “core” methods. As a result of this selection criterion, many methods (albeit sometimes not enthusiastically) were excluded. I do acknowledge the subjectivity involved in any interpretation of the criterion. Nonetheless, I believe the criterion or, perhaps better to say, how it was interpreted, keeps the book highly relevant from a practical perspective and, at the same time, makes it accessible to many learners.

**Скачать Machine Learning with Python: Theory and Implementation**

Пожаловаться администрации