Machine Learning

Ethem Alpaydin

Published by

MIT Press




Reviewed by

Stewart Marshall, MBCS


9 out of 10

Machine learning features heavily at present in both specialist and popular news channels so a book that provides an accessible, hype-free, overview of some of the key ideas in machine learning is a valuable addition to the literature on the subject. Ethem Alpaydin’s book is just such a publication.

This is not a book for those wishing to gain a deep understanding of the theory or practice of machine learning. As an indication of the depth in which it can treat any given subject, the text covers about 170 pages of less than A5 size, but it provides a well written and engaging introduction to some of the concepts and techniques that sit under the banner of “Machine Learning”.

The book starts by discussing why there is so much interest in machine learning and introduces, early on, one of its central tenets: that it is data, not a pre-conceived and hard-coded rule set, that drives the behaviour of machine learning systems. It provides a brief introduction to the ideas behind supervised learning, pattern, face and speech recognition, and natural language processing. There follows a brief but lucid introduction to neural networks and deep learning and an equally clear section on unsupervised learning (clustering and recommendation).

The book concludes with a discussion of the likely trends in machine learning and its application. It is here that the author includes a discussion of some of the social context for machine learning through, for example, comment on concerns such as data privacy and the use of machine learning to control safety critical systems. Particularly topical is the discussion of the narrowing effect that content selection systems based on machine learning might have on the views to which users of news and social media channels are exposed. This section also includes interesting comment on how we perceive and test machine intelligence: we are impressed when computers perform with apparent ease tasks that are difficult for humans but Professor Alpaydin points out that it is more straightforward for a machine to beat a human at chess than for it to recognise the face of its opponent.

This is a brief but thought-provoking treatment of machine learning and will be of interest for those wishing to gain a well-balanced overview of the subject or as a starting point for deeper study.

Further information: MIT Press

March 2017