Author: Nicolas Monmarché
Publisher: Wiley-Blackwell, 2010
Aimed at: Academic researchers
Pros: A useful and up-to-date survey
Cons: Expensive and repetitive
Reviewed by: Mike James
Swarm intelligence is currently fairly trendy - is this textbook a good way in to the topic?
Swarm intelligence is currently fairly trendy and there are some real applications that are worth investigating. This is an academic textbook but with plenty of useful ideas if you want to make a start on practical swarm intelligence.
There is, however, some unnecessary circling around the subject to make it seem more academic and not every reader is going to find it practical enough. The first chapter for example explains many of the basic characteristics of real ants and then extrapolates to what you might want to include in artificial models.
Chapter 2 gets started on the central task - combinatorial problem solving . This is essentially about finding an optimal path and after the basic ideas Chapter 3 presents a survey of the area. From here we look at the problem of optimisation in the continuum and constraint based optimisation.
Chapter 6 on the book takes a case study and practical orientation. There are chapters on the Unit Commitment problem in power distribution, optimisation of aluminium bar production, car sequencing, railway infrastructure, magnetic resonance image segmentation, vertex coloring, hidden Markov models and classification,
The book then moves on to consider less abstract applications - mobile robots inspired by insects and artificial insects. The remainder of the book is a collection of fairly arbitrary topics - network optimisation, detecting organisation, application to disabilities, artificial ant art, ant natural language processing and finally bioinformatics or how to get an ant to fold a protein.
As I was reading the book people did repeatedly ask me if I was serious or was it a spoof? When you read the book you will certainly discover that this approach to AI and optimization if no spoof. However, it is a very narrow technique despite the best efforts of the various authors to convince the reader otherwise. While the assorted case studies are impressive as you read them, what is not clear is how well they could have been solved using other methods. There are no direct comparisons of how effective ants are at finding optimal solutions and no real attempt to either evaluate the approach or link it to other techniques (with the possible exception of the genetic algorithm). Overall the presentation tends to be repetitive perhaps this is indicative of the fact that ant-based algorithms aren't that varied.
In the main this is a suitable book if you are trying to get into the world of artificial ant-based algorithms or just want to see what sorts of things have been achieved. As such it is a useful survey.