Rebala et al. try to explain, in the simplest of terms, the wide diversity of data analysis techniques, that is, machine learning. The book claims to assume no prior knowledge of machine learning, and the authors emphasize a “conceptual understanding of the underlying concepts and the algorithms involved.”

Thus, in chapters 3 to 17, the approaches span a wide horizon: multivariate statistics, classification, clustering, decision trees, neural networks, natural language processing, deep learning, Monte Carlo learning, and many more. Chapter 1 is a rehearsal of basic mathematical and probability constructs, and chapter 2 discusses the different categories of learning/estimation, with only a couple of conceptual examples. Chapter 18 gets a bit more practical with general comments on how to pipeline the design of a machine learning system, from getting data and quality validation to parametric estimation, learning rates, software stacks, and hardware.

From a pedagogical point of view, the volume gives a broad--often superficial--overview of many established approaches. Readers faced with a real data analysis or machine learning case are probably looking for where to start and may get confused when it comes to choosing a method. Readers who successfully implement a technique (or many techniques) on a small dataset will want to understand any errors; the book does not help with this. Prior research shows each technique’s limitations, and it would have been helpful to summarize that information here. A companion volume with sample datasets and solutions, including critical and comparative reviews, would also be helpful.

The book has a detailed table of contents, lists of figures and tables, an index, and a very limited bibliography. It is a significant weakness that many of the presented techniques do not include specific, more thorough references for further reading.

When used as an introduction, start with chapters 2 and 18; then, a quick overview of the techniques in chapters 3 through 17 may flag one to study at the conceptual level. Likewise, the book may be of some help to beginning machine learning software users who want an overview of a specific class of methods. The volume is not suited for real projects or research.