Written by Rich Forestano, firstname.lastname@example.org Wednesday, 30 October 2013 09:31
Sahil Abbi and Arjun Kapoor found out on Facebook that they, along with Connecticut teammate Connor Abbot, were selected as regional finalists in the national Siemens science competition. This team will be among 100 students nationwide competing for regional prizes of $3,000 in November and possibly a grand national prize of $100,000, to be awarded in December.
Kapoor has worked with Stony Brook math professor Dr. Yeufan Deng for the last two years. After his first year, Kapoor looked to continue his research over the summer and reached out to Abbi to meet with Deng and Abbott.
“We all met and we started discussing potential projects,” Kapoor said. “We found this problem that seemed interesting. We had an idea of how we might attempt to solve it and we went from there.”
Going into the summer, Kapoor, 15 of the Wheatley School and Abbi, 14 of Herricks High School wanted to shed light on a method that those in the know, feel there are areas in the field that need tiding up.
“I think what we found most interesting is that yes, there exist the solutions and there was always a way to do it, but it was incredibly inefficient and we thought that would be the best place to start,” Abbi said.
Their project, called “The Optimization Of Parallel Computing Network Topologies Using Simulated Annealing With A Novel Distance Recalculation Algorithm,” focused on the cores of supercomputers. Their group was one of five projects chosen in the New York region as a finalist.
“The way the cores in the supercomputer are connected is really important because without that, you’d have a million small computers instead of one incredibly large, fast one,” Abbi stated. “If you connect them in a regular way, like a grid, it’s not going to run efficiently. You need the best possible way to set it up so that no information is traveling too long or you end up with something called lag time.”
The trio created an algorithm that was capable of “hopping through” a search-base and “converging on the optimal topology for any given supercomputer,” according to Kapoor.
“This makes supercomputers cheaper, more energy efficient and significantly faster,” he said.
Kapoor, Abbi and Abbott wrote an 18-page paper, which most likely be published in a computer science journal, and submitted it to the Siemens contest. The published version will top out at around 40 pages.
From the onset of the project, the team felt their idea would be accolade-worthy, however they wanted to focus on a field and “a desire to look into the field.”
“We didn’t come with the intention of saying ‘Hey, we have to get the best project to win the Siemens,’” Kapoor said. “We thought ‘Let’s try to make a difference. We think we have sufficient knowledge in this field.’”
The team was originally supposed to be aided by graduate students at Stony Brook, but it wasn’t necessary. Deng felt the team proposed solutions to similar problems grad students were working on, so the professor would be the only hands-on mentor.
“For most research projects when you are working over the summer...it’s the grad students project and your calculating a small part of it and whatever data submit is 1/100th of this study that is going on,” Abbi stated. “What Dr. Deng let us do is he gave us the freedom to choose our own path and he gave us the resources and helped us out whenever we ran into a wall.”
The team feels there are still open ended questions in this field, according to Abbi. He said they are still running timing analysis and processing data.
“I think we’ll be trying to close out all this data that we have and I think the peak of it will be a large, nicely-published paper within the next couple weeks,” said Abbi.
Tom VanBell, science coordinator at Wheatley, visited Kapoor and Abbi at Stony Brook and was “blown away” at the work they were doing. He’s taught at Wheatley since 2010.
“It was almost too good to believe,” VanBell said. “I was impressed and happy. I walked out [of Stony Brook] and said to myself ‘Let’s see how this turns out’ and here we are.”
Herricks science coordinator Renee Barcia worked with Abbi, helping edit and review the paper. She has worked at Herricks for six years and called Abbi “one to watch.”
“I’ll be there to help him anyway that I can, to bounce ideas off of me” she said. “I think that Sahil can put his mind to do whatever he wants to do. I feel that he can reach the potential he’s shown and demonstrated already. He has the motivation. He has the desire.”
Abbi and Kapoor are already thinking about their future education. The Herricks standout is researching Harvard, Yale and Princeton as well as Cal Tech. He plans to focus on math.
Kapoor is also looking at Harvard and Princeton but is also researching MIT and The Wharton School of Business at the University of Pennsylvania. His wants to hone his skills in computer science in business or applied mathematics.
All finalists will compete in the regional competition on Nov. 15 and 16. Those winners will advance to the national finals, scheduled for Dec. 7-10 in Washington, D.C.
“At that time, [if we advance], we’d be one of the top team projects in the nation,” Kapoor said.