It is well known that the amount of data available and needing to be processed is increasing worldwide as a result of the widespread use of networked computers in all aspects of society and also the coming of age of new technologies such as the Internet of Things (IoT). A single computing device today possibly has more data storage capacity than all of the computers at a university did a few decades ago, and a person can comfortably carry in a personal device such as a smartphone enough data to fill all of the printed volumes in the Library of Congress.

In this situation, it is also readily apparent that it is no longer possible to work out by hand all the ramifications that may be inherent in a large dataset. Algorithmic solutions must be found to questions like: Is a purchaser of a certain item in an online store also likely to be interested in some other item? What do the data show about the efficacy of certain public policies? If a government were to reduce the income tax and increase the sales tax on certain goods, would this increase the GDP or the net revenue? How can law enforcement agencies find the hidden behavioral patterns unique to individuals who may be part of sleeper cells of terrorist organizations?

The field of machine learning finds its application in precisely this context. Built out of a judicious mix of statistics and algorithms (with appropriate use of discrete mathematics, linear algebra, operations research, and so on, added where needed), it enables us to find ways to gain knowledge from massive datasets that cannot be made sense of otherwise.

This book is the third edition of a well-received work that does a good job of covering, in a single volume, a sampling of the spread of machine learning techniques in a manner suitable for a beginning student (who may be an advanced undergraduate or an early graduate student). The book is also usable by a practitioner who may wish to understand and use some specific machine learning techniques. In 19 chapters, including one of general introduction, the author covers the basics of important machine learning techniques such as supervised learning, reinforcement learning, multivariate methods, clustering, dimensionality reduction, hidden Markov models, and so on. An appendix gives some relevant background in probability theory, and the exercises at the end of each chapter are a welcome resource for instructors and students alike.

However, as the focus of the book is on the mathematical concepts rather than on programming and actually working with datasets, practitioners or academics who wish to do hands-on work using machine learning will probably have to find other resources to help with such endeavors. Some theoreticians and practitioners may also find it useful to blend their study of machine learning with some exposure to classical statistics-based approaches, such as those in the books by Spirtes et al. [1] and Pearl [2].