Home About Articles Graduate students win prestigious data mining competition

Graduate students win prestigious data mining competition

August 29, 2007

Azencott and S. Josh Swamidass, two graduate students in computer science professor Pierre Baldi's lab, finished in first and second place respectively in the data mining competition "Agnostic Learning vs. Prior Knowledge" part of the International Joint Conference on Neural Networks (IJCNN) 2007 Conference, the premier conference in the field of neural networks.

The challenge revolved around 5 data sets from various domains, including marketing, ecology, text classification, handwriting recognition, and drug discovery.

Azencott and Swamidass focused on the chemoinformatics drug discovery HIVA data set consisting of 42,678 compounds tested for activity against the AIDS HIV virus.

Of these compounds, 4,229 were labeled "active" or "inactive", forming the training set.

A total of 45 teams participated in the overall data mining competition, 30 of which participated in the HIVA portion of the contest.

A pre-run of the competition was held from October 1st, 2006 to March 1st, 2007, while the final portion of the contest ran from March 1st, 2007 to August 1st, 2007. Final results were published online.

Azencott was invited to present her results at the IJCNN07 Agnostic Learning vs. Prior Knowledge Competition and Data Representation Discovery workshop, August 17th, 2007, in Orlando, Florida and was awarded a monetary prize.

The top results were obtained using Support Vector Machines, with 2D kernels specifically designed for small molecules in Baldi's lab.

Azencott is entering her third year as a graduate student in ICS. Swamidass is an M.D./Ph.D. student who obtained his Ph.D. in ICS in June 2007 and is now in the process of completing his M.D. degree at UCI.