Math Modeling Contest Adds Up to Major Achievement for Wooster Trio
Students finish in top one percent after immersing themselves in network modeling
WOOSTER, Ohio — Just hours before dawn on a frigid Monday morning in February, a trio of students from The College of Wooster experienced a Eureka moment. After spending 80 hours holed up in a classroom turned topsy-turvy on the third floor of Taylor Hall, sophomores Sayantan Mitra, Dagmawi Zegeye, and Paroma Palchoudhuri finally figured out a solution to a problem posed by the Consortium for Mathematics and its Applications 16th annual international Interdisciplinary Contest in Modeling (ICM).
The four-day competition involved data analysis, network science, and mathematical modeling. The Wooster contestants chose a problem that required them to describe the nature of influence and its impact in a citation network and co-author network using a model, and further applying that model to other fields of life, such as social networks. In other words, it was complicated.
Nonetheless, the three Wooster students, who had been together in Mary Jo Kruezman's Transition to Advanced Math class, took on the challenge. "I asked 'Popi' (Palchoudhuri) and 'Dagm' (Zegeye) if they would be interested (in the contest)," said Mitra (a.k.a. "Sunny"), an economics and mathematics double major from Calcutta. "We really didn't know what we were getting into."
Indeed, they were about to embark upon a four-day odyssey that would be both maddening and exhilarating. They entered Room 308 of Taylor Hall at 8 p.m. on Thursday evening, and with the exception of a few outside meals and periodic bathroom breaks, they did not leave until 8 p.m. on Monday night. They worked, ate, slept, and occasionally bickered for the next 96 hours, but when it was over, they had come up with something quite remarkable. Their research, summarized in a 20-page paper, placed them in the top one percent of all entries for that particular problem (11th of 1,028 teams), netting them a designation of "Finalist."
"We didn't do it as a competition," said Palchoudhuri, a physics and mathematics double major from Calcutta. "We did it for the experience."
And what an experience it turned out to be. The three quickly divided responsibilities based on each individual's area of expertise. Mitra reviewed the literature and conducted the research; Zegeye handled the programming and model development; and Palchoudhuri did most of the writing. They were essentially on their own while trying to work together, which was not easy. "We tried to stick together for 96 hours without going crazy," said Mitra. "At one point we even considered opting for another question."
Ultimately, the collaboration proved to be successful, and the trio managed to come up with a model, but they had trouble simulating it on the computer. "One of the initial hurdles we faced in developing the model was extracting the data provided," said Zegeye, a mathematics and computer science double major from Ethiopia. "It was a big dataset, and we needed to parse through it to remove bits we didn't need and create a network representation of it.
"We were also asked to choose our own dataset," added Zegeye. "We wanted to analyze something that other teams perhaps wouldn't look at, so instead of analyzing a network of film actors or Facebook friends, we chose to examine a network depicting international conflict (a war network)."
Once the data was ready, the trio was faced with the challenge of developing an algorithm that was capable of measuring the influence of any entity within the network. This would also allow them to find the most influential entity.
Finally, at 4 a.m. Monday their persistence paid off. They came up with a model, and four hours later, they were able to simulate it on the computer. That gave them 12 hours to explain their methodology, summarize their findings, outline their solution, and write their paper.
"We investigated three different networks: an Erdös co-authorship network, a citation network of foundational Network Science, and an international conflict network," said Zegeye. "Our paper presents the model we developed (based on Katz centrality) and its possible applications."
The trio had to wait almost two months for the results, but when the evaluation was complete, the judges had ranked the Wooster team ahead of any other U.S. team and in front of thousands of international competitors, many from China.
"It was quite an accomplishment," said Matthew Moynihan, visiting assistant professor of mathematics and computer science. "They did great work in an extremely compressed time frame. Sunny and Dagm are both in my Graph Theory course this semester, but the competition came so early in the term that they were still basically working from scratch. Sunny, Popi, and Dagm are all double majors, and that definitely helped them look at the problem from different angles. The competition was a great introduction to research: tedious, frustrating, and exhilarating all at once. Who knows, perhaps this will lead to an I.S. (Independent Study — Wooster's nationally acclaimed undergraduate research experience) project in a few years."