Features
The Origin of a Computer Scientist
U of G prof aims to improve tools whose principles evoke Darwin's theory of natural selection
BY ANDREW VOWLES
 |
| Prof. Mark Wineberg studies computing principles borrowed from the natural world. Photo by Martin Schwalbe |
It wasn't the HMS Beagle of Darwin fame but a converted freighter on its final tour of the Galapagos Islands before being retired. And although Prof. Mark Wineberg had grown up as a backyard stargazer at home in North York, he was hardly a naturalist in Charles Darwin's league. But a trip Wineberg took to South America in 1990 evoked something of the voyage taken more than a century and a half earlier that ultimately led the British naturalist to his theory of natural selection.
It helped that the then recent B.Sc. graduate from the University of Toronto had taken along a propitious choice of reading material.
I thought, if I'm going to the Galapagos Islands, I might as well read the Origin of Species, says Wineberg, who a decade later would pack up his belongings for a move to Guelph. Today he's a faculty member in the Department of Computing and Information Science, where his efforts to improve computing draw on principles borrowed from the natural world.
Recalling his arrival at U of G in 2000, he says: What attracted me to Guelph is that this department has people doing work in natural computation. Natural computation takes processes observed in nature and applies them to solve complex tasks.
For several of his colleagues studying neural networks, that means using the human brain and its processes as an analogue for the computer. But Wineberg's interests lie in a branch of natural computation using genetic algorithms in evolutionary computing. Genetic algorithms offer a kind of search technology based on optimization, or looking for the best solution from among a number of candidates.
Optimization is well-known to engineers and managers charged with designing processes and schedules to run as efficiently as possible. How to set up machines on the factory floor to make different widgets without having to retool for every batch? How to organize the job shop schedule to make efficient use of resources, human and otherwise? Elsewhere, managers might be looking for optimum prices and quantities of goods, not just at one time but also as market conditions change.
Consider potential solutions as a population of points that continually change. And look at each solution as a string of information akin to genetic code, also open to change or mutation. Although greatly simplified, that model mirrors the basic elements worked on by natural selection as living things adapt to their environments. And it offers a way to use computing to solve complex problems in a dynamic environ- ment, says Wineberg.
(The analogy isn't entirely artificial. He points out that your eye is a solution to the problem of vision, but it's only one of a number of possible solutions that natural selection could have worked on to enable us to see or otherwise sense our sur- roundings.)
Rather than apply genetic algorithms directly to real-world problems in, say, manufacturing or business, Wineberg straddles theory and practice. He models genetic algorithms to better understand and refine them. Improve the tools, he says, and perhaps others will find ways to apply them in tomorrow's computers and computing environments.
The same ideas that use computing to solve problems may also help improve computers themselves. A subset of genetic algorithms called genetic programming, for instance, is used in machine learning and in designing computer chips. And turning the natural computing analogy back on itself, genetic algorithms might also improve bioinformatics biologists' use of computing power to analyze mountains of data and solve biological problems.
At Guelph, Wineberg is working with a geography professor using multi-objective genetic algorithms to solve a PhD problem: How to optimize a farmer's crop yield with judicious use of chemicals to avoid environmental contamination?
He completed his master's at Carleton University in 1993 after switching from parallel algorithms to evolutionary computing. He was inspired to pursue a doctorate after auditing a fourth-year course in population genetics at Carleton. Earlier, the longtime science fiction fan had studied physics, computer science and astronomy at the University of Toronto.
He found himself drawing on his undergraduate studies and his boyhood passion for the stars during that tour of the Galapagos Islands in 1990. In a kind of trade-off, he learned about the biology and geography of the region from the on-board naturalists, then helped them make sense of the stars populating the equatorial night sky.
And between times, he found himself returning to that seminal volume he'd brought along.
It's more of a spiritual connection, sitting on the deck of the ship reading the Origin of Species in the Galapagos Islands, he says. I can't explain it it was just a connection. You're reading the book in the place where the book had in some respects been inspired.