Thursday, 7 July 2016

Information Theory: a tutorial introduction - James V. Stone ***

Information theory is central to the technology that we use every day - apart from anything else, in the technology that brings you this review (though ironically not in the book being reviewed as it doesn't appear to have an ebook version). As in his Bayes' Rule, James Stone sets out to walk a fine line between a title for the general reader and a textbook. And like that companion title the outcome is mixed, though here the textbook side largely wins.

The opening chapter 'What is information?' walks the line very well. It gradually builds up the basics that will be required to understand information theory and though it would work better if it had a little more context (for example, more about Claude Shannon as a person) to anchor it, the general reader will, with perhaps a few pages that needs re-reading, find it approachable and providing more depth than a popular science title usually would. I like the way that Stone uses variants of a photograph, for instance, to demonstrate what is happening with different mechanisms for compressing data. Unfortunately, though, this is pretty much where that general reader gets off, until we get to chapter 9.

The main bulk of the book, pages 21 to 184, cross that line and plonk solidly into textbook territory - they may cover the topic rather more lightly than a traditional textbook, but they simply don't work to inform without requiring the kind of investment of mind and mathematics that a textbook does - and, with a few brief exceptions, the writing style feels no different from the better textbooks I have from university. At chapter 9, the subject is brought round to information in nature, and there we get enough application and context to make what we learn seem more approachable again, though not to the same level as the equivalent part of Bayes' Rule. It's also a shame that (unless I missed it) there is no mention of Omega, Greg Chaitin's remarkable non-computable number.

So where Bayes' Rule is suited to popular science readers who want to stretch themselves and put in some extra effort, Information Theory can only really be regarded as a readable introductory textbook - it doesn't work in a popular science context. (Why then am I reviewing it? The author kindly provided the title for review in the hope that it would work for popular science readers.) If you are about to take a university course encompassing information theory - or are contemplating doing so - I can, however, heartily recommend this title as an introduction. 


Buy direct from the author here.
Review by Brian Clegg

No comments:

Post a Comment