Kelly '21 Develops Genetic Programming
“Computers are going to take over the world. This is step one.”
Peter Kelly ’21 loves his summer computer science research so much that he will explain genetic programming three times over and never lose his smile. He is both awed and humbled by the technologies he is researching and has ardent faith in the developments it can bring about.
Kelly’s genetic programming research, under the direction of Assistant Professor of Computer Science Tom Helmuth, evolves computer programs to try to solve a problem. In genetic programming, the user has a problem or a puzzle for the computer to solve. The computer generates random programs that attempt to solve the problem, eliminating the programs that do not solve it, and then the programs that work well –“the parents” – combine to create new programs that work better.
Helmuth’s project for Kelly focuses on a type of parent selection (lexicase selection), which randomizes the order of tests on which each program is evaluated for selection, ultimately creating programs more efficient at solving problems than standard tournament-style genetic programming.
Kelly’s job is to create a “benchmark suite” which comprises problems for genetic programming to solve. Helmuth already has a benchmark suite of 29 problems that he has used to test his work, but the problems are all taken from computer science textbooks and are occasionally problematic. These problems include basic introductory problems, such as asking a computer to count the number of vowels in a sentence.
Major: Computer science
Hometown: Tacoma, Wash.
High School: Glastonbury High School
Kelly will be spending his research time going through textbooks and searching online to find problems that can potentially be solved by genetic programming but also prove somewhat of a challenge for the programs to solve. His updated benchmark suite will altogether help demonstrate the bounds and capabilities of genetic programming and other automatic program synthesis techniques.
“So, right now, genetic programming is still fairly new. It’s just going to get bigger,” Kelly said, explaining how his current tests may seem basic but will eventually advance and require more from the programs. He also said that genetic programming’s novelty is one of the reasons he is excited to work on the project. “I truly think that what I’m doing is one of the coolest things ever.”
Kelly also said that he enjoys working on the project because of the independence Helmuth grants him. Explaining how his adviser allows him to set his schedule and research plans, he said, “I think there’s a lot of great professors. Helmuth’s the best one — just saying … He trusts us.”
Hoping to eventually be the project lead of a video game, Kelly cites Hamilton as a place that will help him achieve his aspirations. “I think with Hamilton, the fact that I can write well and have good communication skills as well as the science side is great,” he said, praising the liberal arts approach to education.
Kelly can’t wait to see both where his education takes him and how genetic programming advances, and his summer research with Helmuth ultimately acts as excitement and inspiration for the future. “It’s going to become big.”