Wednesday, November 30, 2016

URSP student Anson Rutherford Studies Genetic Algorithms to Creatively Solve Problems

My project is focused on trying to use genetic algorithms to solve problems in creative ways. Genetic algorithms are used to solve a problem where we can judge the success of an individual “guess”, but don’t have any way of predicting what guesses will perform well before testing them. A typical genetic algorithm will make a number of random guesses and use the results of those guesses to better inform future guesses, often by making guesses similar to those that performed well. Given enough time, a genetic algorithm will find an optimal solution. My research hopes to adapt the many great things these algorithms have to offer to problems where we want to find more than a single optimal solution. Instead, I hope to construct an algorithm that can find a wide variety of acceptable solutions to a given problem. This is useful is situations where we’re more interested in learning about the question than in solving the problem.

Genetic algorithms are inspired in a lot of ways by biological evolution, which was one of the reasons I became interesting in genetic algorithms in the first place. A lot of my initial fascination came from experiments that simulated aspects of biological evolution, evolving either the physical structure or intelligence of an artificial organism over time. Finding an overlap between my interests in evolution and computer science was exhilarating, and I have been studying the subject ever since. I have always enjoyed learning for its own sake, and hope to meld these three interests into a career of research and discovery.

My project requires me to simultaneously plan and implement the algorithm. Obviously a lot of coding is involved, and so I’ve spent a lot of time programming basic functionality so that we can do build off of it. A lot of that involves implementing similar systems and testing their functionality. The other aspect of my project has been researching the many other genetic algorithms that exist, and using that information to form my own. Learning about the innovations of other researchers has been fascinating, and I hope that my research can help add to the wealth of useful information on the subject.