This is the era of machine intelligence. The domain of machine learning, and in particular deep learning, has evolved rapidly, and we have started to develop and use real-life applications from this evolution. The book serves the purpose of providing an introduction to deep learning models and realizing them using the programming language R.

The first chapter provides a quick overview of the landscape of deep learning, including models such as convolutional neural networks (CNNs) and recurrent neural networks (RNNs). Mathematics is an essential element of machine learning methods; keeping this in mind, the author covers the essentials of mathematics in chapter 2. It includes statistical and linear algebra concepts. Chapter 3 discusses the basics of machine learning models, including regression models and unsupervised learning models such as k-means clustering. This chapter also discusses the trickier aspects of machine learning such as choosing an appropriate learning rate. The book dedicates individual chapters to deep learning models, from perceptrons to CNNs, RNNs, and even autoencoders and deep belief networks (DBNs). The last two chapters use R to implement some examples.

Providing a quick and high-level overview helps readers understand the complex domain of machine learning. One of the book’s strengths is its discussion of some tricky aspects of machine learning, for example, hardware considerations, experimental design and heuristics, and choosing an appropriate learning rate.

Despite some good aspects, the book leaves a lot to be desired. It lacks rigor and flow, that is, the rationale for what is being presented is often unclear. For example, the motivation to use kernels in support vector machines (SVMs) is not discussed, leaving the reader clueless. Similarly, the structuring of concepts is far from appropriate, for instance, logistic regression and SVMs are subtopics under “Testing for Multicollinearity,” which does not seem relevant. Another concern is related to the treatment of mathematical formulas. Mathematical expressions and formulas are expected in such a book; however, many such formulas are presented without any explanation or rationale (such as one mentioned in a section on the pooling layer in CNN in chapter 5). Further, a reader expects the application of theoretical concepts to be realized into practical examples in the same chapter. In general, the author did not put sufficient effort into the examples; specifically, no example is included in the chapter describing CNN. Lastly, the book has many typographical errors and the figures are not typeset well (some are stretched, some look very big, and some are too small to read).

The book provides an overview of deep learning techniques and explains how to use R to design and implement a deep learning model. It may be a beneficial read for beginners interested in designing deep learning models in R.

More reviews about this item: Amazon, Goodreads