SLIPPERY ROCK, Pa. - A five-member team of Slippery Rock University computer science majors was awarded the "Best Interview" title at the recent Supercomputing Conference hosted in Salt Lake City.
The team, which spent two-and-a-half months preparing for the conference's cluster competition, included: Carter Danko of Canfield, Ohio; Robert Lindenberg from Delmont; Kyle Messner from New Bloomfield, James Mullen from Franklin; and Michael Tasota from Monroeville. David Valentine, SRU professor of computer science, accompanied the team.
The award is given to the team demonstrating to a panel of judges the most thorough comprehension of parallel computer architecture and programming.
For 24 years, SC12 has been at the forefront in gathering the best and brightest minds in super-computing together, with unparalleled technical papers, tutorials, posters and speakers, Valentine said.
The cluster competition was hosted in the conference's Exhibiters Hall beside top-ranked hardware vendors displaying the latest in high-performance computing gear. Exhibitors included Intel, Nvidia and Microsoft among others. "Several vendors asked our team for resumes or contact information," Valentine said.
"The SRU students set up the hardware, tuned the software stack and generated an original program to solve the 'Traveling Salesman Problem,'" Valentine said.
Danko said the team "had been reading up online about edge assembly crossover algorithms before the competition and saw a way to employ them in solving the problem."
"The interview portion of the competition was with some of the leaders of the conference and since the team has been active, we were knowledgeable and knew what was going on and how to be involved," he said.
"The TSP problem is very easy to describe, but very difficult to solve: A salesman must visit 'N' cities and return to his base without duplicating cities. The students are required to find the shortest route for his tour," Valentine said.
"The competitors were given multiple datasets of 10,000 cities to solve the problem. A brute force algorithm could run for days on such a dataset before finding the 'best' solution," he said. "The team had to find 'good' solutions for 10 different datasets, and all within a 48-hour contest window."
The team elected to use a "genetic algorithm" to attack the problem. "GA's treat the solution like a set of 'chromosomes with various genes' and seek to find 'better' solutions by combining [mating] good solutions in an evolutionary approach," Valentine said.
"For me, one of the highlights for SRU was the team's 'visualization' of the data. They were the only team to display a video representation of each dataset. Seeing their visualization, the judge creating the datasets started playing games with the data. We started to see 'pictures' such as the Mona Lisa, the SC12 logo, etc., in the datasets," Valentine said.
"It was fun to see other teams get the data, start their code and then run to our booth to see what it looked like," he said. "Because of their work, the SRU team directly shaped the fundamental content of the contest. And of course, winning the interview phase of the contest is a real feather in their caps."
Danko said he made contact with a number of vendors and conference participants - Purdue University and the Los Alamos National Laboratory - and hopes one of them might lead to an internship.
"The last dataset released was a simple, regular rectangular grid of 10,000 points. There is an obvious, brute force solution to such data, and SRU was the only team to recognize it - because of their visualization. They were able to quickly generate a new bit of code and submit the resulting optimal solution. We can proudly say that only SRU generated a fully optimal solution for any of the datasets," Valentine said.
"Our classwork before the competition using the genetic algorithm just seemed appropriate once we saw the competition problem," said Tasota, who graduates next spring. "We had thought the problem was going to be totally different, but once we saw it, we knew the algorithm should work."
"I think everyone found the visualization very appealing and we were able to watch it develop on our booth monitor," he said.
Dan Olds, a reporter for the British hi-tech ezine The Register, covered the SC12 and has posted video of his interview with the SRU team. It is available at: http://www.theregister.co.uk/2012/11/14/sc12_cluster_competition_vids/.
On their return to campus, the team presented a summary of its work to the SRU Computer Science Department Board of Visitors.
Slippery Rock University is Pennsylvania's premier public residential university. Slippery Rock University provides students with a comprehensive learning experience that intentionally combines academic instruction with enhanced educational and learning opportunities that make a positive difference in their lives.