Joe Mullich
Freelance Health Writer
818-907-9109
Advertising Age
“Survival of the Fittest” Software
By Joe Mullich
For three-and-a-half billion years, nature has used a nifty trick for improving itself: It subtly mutates the genes in organisms, letting the strong attributes survive and the weak ones perish. The Darwinian “survival of the fittest” theory is now being embraced by computer programmers. Software incorporating genetic algorithms (GA) mimics evolution: Lines of computer code act like living organism, continually interacting and mutating with each other. The software evolves, leaving only code that is working toward the optimal solution of a complex problem.
GA is now being used for such things as managing financial portfolios, designing communication networks and improving manufacturing schedules. Advanced Geophysical Co. in Denver, Colorado, includes a GA optimization technique in its software package to interpret seismic data. LBS Capital Management System in Clearwater, Florida, picks stocks for pension managers with the aid of GA. Moody’s Investors Services, a bond-rating agency in New York, uses GA to schedule hundreds of jobs for supporting the 1,000 computers spread throughout its facility. Texas Instruments used GA to design a computer chip that required 18 percent less space than the best design that human minds could conceive.
While the concept behind GA is as old nature, the marketplace for this technology is embryonic -- but growing. “In the past five years, we’ve gone from maybe a dozen GA applications to hundreds of applications today,” said Lawrence Davis, president of Tica Technologies Inc., a consulting firm in Cambridge, Mass. that specializes in GA.
GA were invented in the 1960s by John Holland, a professor at the University of Michigan who received a MacArthur Foundation “genius grant” for his efforts. Professor Holland realized the ones and zeros that were strung along a computer code bore a stark similarity to the way genes are placed along a chromosome. In evolution, organisms pass on mutant genes to their offspring, and the offspring which are improved by the new genes tend to survive. This survival of the fittest technique can be applied to computer code by rearranging the ones and zeros. For more than a decade, though, GA remained hiddened in universities and research centers.
That began to change when new techniques and programming languages, like Windows and Visual Basic, made GA easier to use. That promise could be enhanced even more by the Java programming language. Some of the most recent GA applications require no programming ability, creating the beginnings of a GA marketplace.
GA is an optimization tool -- it helps other things do their job better. When a computer is not being used for anything else, GA will automatically and transparently attempt different scenarios, calculating and crunching data to find better solutions. While a programmer is getting a cup of coffee, a GA can allow his computer to explore different variables, pieces of codes, or correlations in data that will make a program or procedure operate more efficiently.
“GA should only be a small cog in a big machine that makes the machine run better,” said David Orvosh, a programmer at Tica Technologies. “The trick is figuring out what products you can create to make the genetic algorithm a market-differentiating feature. It has to have a big payoff.”
True enough. A genetic algorithm could be used to, say, optimize the design on a plant floor, allowing three more products per hour to be produced. But that might not be worth the trouble of repositioning and purchasing lots of heavy equipment. For high-tech design projects, like designing fiber-optic networks, the ante can be enough to warrant the use of a GA. In the past, US West, a Baby Bell and Tica Technologies client, let designers lay hugh networks of fiber-optic cable according to their instinct and experience. Using a GA that kept crunching data to find networks that needed the least amount of cable, US West reportedly cut design times for networks from two months to two days, saving up to $10 million per job. One of the hottest areas for GA now is the financial industry, where incremental improvements can translate into enough dollars to encourage experimenting with new technology.(continued)
1 2