Evolutionary Algorithms Make Faster Design a Reality

August 13, 2007

Thanks to powerful computers and mathematics, evolutionary algorithms can create new designs and improve upon older applications faster than ever, according to New Scientist. Evolutionary algorithms mimic the processes of natural selection by "breeding," selecting, and recombining possible designs to produce the "fittest" ones.

In producing improved designs, evolutionary algorithms can test and combine large numbers of features in ways that might not occur to human designers. "Human engineers usually design stuff by tweaking a few parameters," Steve Manos of University College London told New Scientist.

Moreover, the new generation of evolutionary algorithms can handle design problems previously considered too complicated or time-consuming to tackle. "Things we couldn't have done in the past, because it would have taken two months to run the genetic program, are now possible in days or less," said John Koza, a computer scientist and genetic algorithm pioneer at Stanford University.

To encourage even greater use of evolutionary algorithms, the Special Interest Group for Genetic and Evolutionary Computation of the Association for Computing Machinery sponsors the Human-Competitive Results awards, which single out designs that are "competitive with the work of creative and inventive humans." This year's winners were announced at the Genetic and Evolutionary Computation Conference (GECCO 2007), in London, in July.

Steve Manos won the $5000 gold prize for using evolutionary algorithms in the emerging field of "holey" optical fibers. Another was a prize winner by demonstrating fast work. Pierre Legrand and colleagues from the University of Bordeaux 2, France, took barely a day and a half to reconfigure the electrodes in cochlear implants to create an optimal pattern for one patient whose doctors had not succeeded in doing so in 10 years!

Detractors have suggested that the mathematics underlying some evolutionary algorithms is intractable. "If you don't know how an evolved design works, how can you know when it might fail?" they ask. But Koza called the objection "self-serving and bogus." To overcome such objections, Koza noted that "you can test the hell out of the one solution you settle on."

Source: New Scientist, July 28, 2007.