How to Solve It: Modern Heuristics

by Zbigniew Michalewicz, David B. Fogel

No pleasure lasts long unless there is variety in it. Publilius Syrus, Moral Sayings We've been very fortunate to receive fantastic feedback from our readers during the last four years, since the first edition of How to Solve It: Modern Heuristics was published in 1999. It's heartening to know that so many people appreciated the book and, even more importantly, were using the book to help them solve their problems. One professor, who published a review of the book, said that his students had given the best course reviews he'd seen in 15 years when using our text. There can be hardly any better praise, except to add that one of the book reviews published in a SIAM journal received the best review award as well. We greatly appreciate your kind words and personal comments that you sent, including the few cases where you found some typographical or other errors. Thank you all for this wonderful support.

Genres: Mathematics, Computer Science, Science, Reference, Nonfiction, Programming, Computers

570 pages, Hardcover

First published January 1, 2004

Reviews

[Barry · Rating 3 out of 5]

This is a really annoying book.

Yellow Springer books are generally not intended to be read cover-to-cover. This one was approachable enough that I could (and did). I was already familiar with many of the problems and algorithms presented. The bibliography is quite complete, and even if the index is a little skimpy, I've used is as a reference book.

Basically, it's a high-level overview with lots of tantalizing results and pointers to the literature. If you're in CS, it's probably worth picking up.

But skip the brain teasers.

Just trust me on this, ok? They're awful.

Here's a hint from the "What are the ages of my three sons?" puzzle on page 9.

"....That's fine" says the mathematician, "but you'll have to tell me something about them."
"The sum of their ages is equal to the number of windows in that building," says the father, pointing at the structure next to them.
....
[To solve the puzzle,] [w]e have to assume the mathematician knew the number of windows...

We do? I didn't. And the problem can't be solved without that assumption. Grrrr.....

All of the brain teasers are like that. And, believe it or not, that's the biggest impression the book made.

[Ushan · Rating 3 out of 5]

A collection of heuristics for solving difficult (NP-complete etc.) problems approximately. You learn most of this material in an artificial intelligence/combinatorial optimization/machine learning college course. One topic I wasn't familiar with is coevolutionary genetic algorithms. If you want to evolve rabbits that excel in escaping from being eaten by foxes, you should also at the same time evolve foxes that excel in catching rabbits. For example, if you want to evolve parallel sorting networks, you should also evolve lists of numbers that the networks sort incorrectly. One of the authors evolved checkers-playing neural networks that played against each other; the winner was registered at MSN Games as Blondie24 and made it to the top 500 out of the 120,000 checkers players registered at the site.

[Jay Reynolds · Rating 3 out of 5]

Well written and clear, 'How to Solve It' is a condensed collection of approaches to numerical optimization, and explains their applicability to various problems. Both gradient based and gradient free methods are presented. There's really not anything in this book that can't be found on the internet, but I find it fun to flip through occasionally, nonetheless. - a good book for anybody (you needn't be a mathematician) interested in browsing a diverse sampling of optimization techniques.

[Patrick]

I've read bits and pieces from this book. It's not as interesting as I originally thought it would be, sadly.

[Nata]

Actually pretty good but couldn't finish it since I had to return it. Definitely will pick it up again in the future.