Multi-Objective Optimization Using Evolutionary Algorithms

Rating:
Summary: Great book; a must for engineers and scientists alike
Review: Kalyanmoy Deb has put together a great summary of the state of affairs in multiobjective genetic algorithms. Should you be an engineer or a scientist involved in the optimization of any design of sizeable complexity, you should read this book and become familiar with the techniques that have evolved over the last decade into powerful methods of optimization. This book is in many many ways bridging the gap from Michalewicz's and Fogel's book ("How to solve it") to the more modern era of this field (eg late nineties up to now...). So whereas those two authors never really considered multiobjective genetic algorithms, Deb plows through with the great expertize of a (perhaps even "the") leading researcher in that domain. This is a great book of _receipes_ with the level of details necessary to make use of them. It's a "how to" book; this is the one you have cracked open on your desk while you're hard coding it all up. However, it's not very well written with the prose being very terse and basically quite unengaging. But so what! In some sense yes perhaps, but Michalewicz and Fogel made a point that one can write technical litterature that one can also read. Perhaps they went overboard... in any case, Deb's book is about algorithms so who cares about whether the book puts you to sleep and it can do that, unfortunately. Apart from the unengaging style and the paucity of depth in the examples scope, the real problem with the book is not with the book itself, it's with the field of multiobjective optimization based on evolutionary methods. It's fairly evident that there is not much of any sort of fundamental understanding available at this time in support of why evolutionary techniques do work well, and they do, sometimes... If this understanding is available, you won't find it in Deb's book. If you are like me though, you won't care all that much really so long as the techniques are efficient and presented in a way that make them useable, and that's done right... But on the whole, it's a little unsatisfying because one's left with a panoply of various techniques and ways to define operators and representations but there is no insight given on which one might be best or how to craft them to particular situations. There is a lot of so-'n-so in reference this and that did it like this and it seems to work well there, however... The reason for this state of affairs is, of course, that nobody has a real clue, yet... But that is _not_ Deb's fault and this is not why, as a user, I'm not rating his book a full 5 stars. In some sense it could be rated as high as that but I thought the presentation was rather unengaging and not with all the breath and depth it could have had. So it's a 4.5 stars perhaps... let's say... but Amazon does not let me select 4.5 stars so it's 4, this edition at least...

Rating:
Summary: Great book; a must for engineers and scientists alike
Review: This is the first complete and updated text on Multi-objective Evolutionary Algorithms (MOEAs), covering all major areas clearly, thoughtfully and thoroughly. Thanks to the development of evolutionary computation MOEAs are now a well established technique for multi-objective optimization that finds multiple effective solutions in a single run. The widely interdisciplinary interest of engineers, scientists and mathematicians towards MOEAs has been evident during the first international conference on this topic (EMO2001,Zurich). The book is extremely useful for researchers working on multi-objective optimization in all branches of engineering and sciences, that will find a complete description of all available methodologies, starting from a detailed description and criticism of classical methods, towards a deep treating of the most advanced evolutionary techniques. Moreover several analytical test cases are given, covering all difficulties a MOEA encounters when converging towards the Pareto Optimal front. This set of test problems, together with several performance measurement parameters are essential when testing a new strategy before its application to a real-world problem. Despite the detail in advanced topics, Deb's book may be also used as a reference-book for a post-graduate course thanks to the scholarly coverage of basic arguments. As a final remark I strongly suggest everyone working on evolutionary computation and optimization to keep this book on the desk.

Rating:
Summary: The Reference in Evolutionary Multiobjective Optimization
Review: This is the first complete and updated text on Multi-objective Evolutionary Algorithms (MOEAs), covering all major areas clearly, thoughtfully and thoroughly. Thanks to the development of evolutionary computation MOEAs are now a well established technique for multi-objective optimization that finds multiple effective solutions in a single run. The widely interdisciplinary interest of engineers, scientists and mathematicians towards MOEAs has been evident during the first international conference on this topic (EMO2001,Zurich). The book is extremely useful for researchers working on multi-objective optimization in all branches of engineering and sciences, that will find a complete description of all available methodologies, starting from a detailed description and criticism of classical methods, towards a deep treating of the most advanced evolutionary techniques. Moreover several analytical test cases are given, covering all difficulties a MOEA encounters when converging towards the Pareto Optimal front. This set of test problems, together with several performance measurement parameters are essential when testing a new strategy before its application to a real-world problem. Despite the detail in advanced topics, Deb's book may be also used as a reference-book for a post-graduate course thanks to the scholarly coverage of basic arguments. As a final remark I strongly suggest everyone working on evolutionary computation and optimization to keep this book on the desk.