Browse Books

Reviewer: Denilson Barbosa

Understanding machine learning is a most welcome breath of fresh air into the libraries of machine learning enthusiasts and students. Unlike all other previous texts, this book dives deep into the theory first, looking at foundational and hard questions, before moving on to specific algorithms. The narrative, always formal and sometimes terse, uses illustrative and fairly intuitive examples to get the message across very effectively. Proofs are explained in detail, and each chapter ends with a good list of exercises. The book starts with a brief and crisp philosophical discussion framing the subject of the book, statistical learning from data, and then dives into foundational questions: What is learning__?__ What can be effectively learned__?__ At what cost__?__ These issues are discussed in the first few chapters of the book, which provide a formal model of learning (the classical probably approximately correct (PAC) model), and a universal learning framework. Also in this introductory section, other often-ignored yet foundational concepts, such as the no-free-lunch theorem and the computational cost associated with learning, are rigorously discussed. Next, the book devotes quite a lot of space to the wealth of successful learning methods in the literature, covering them in light of the theoretical background laid out in the first section. As a result, the exposition is crisp, taking the reader immediately to the nitty-gritty of the matter. All familiar methods are covered, from linear predictors to kernels, support vector machines, decision trees, and neural networks, as well as other popular topics such as boosting, gradient descent, multiclass, and ranking problems. The book has two other sections: one covering additional topics and the other delving deeper into the theory. Among the additional topics, the chapters on clustering, feature selection, and dimensionality reduction were particularly interesting. In summary, this is a thorough and very well-crafted textbook, aimed at graduate students and researchers, but accessible to senior undergrads as well. It clearly synthesizes the field in a much more elegant way compared to previous books, which often are essentially collections of research papers edited into fairly disconnected chapters. As such, it is a more than welcome addition and a very strong contender to be the standard textbook in machine learning for years to come. More reviews about this item: Amazon Online Computing Reviews Service