Back in May 2014, Distinguished Professor of Computer Science Pierre Baldi and his Ph.D. student Peter Sadowski published a paper, “The Dropout Learning Algorithm,” in Artificial Intelligence Journal (AIJ). Five years later, the paper has been recognized with the 2019 AIJ Prominent Paper Award. The award recognizes outstanding AIJ papers — published within the past seven years — that are “exceptional in their significance and impact.”

“Dropout is an important machine learning algorithm, used primarily to avoid overfitting when training large neural networks,” explains Baldi. Avoiding overfitting is critical to developing models that generalize well, far beyond their training data. “This paper was the first to derive a mathematical theory of dropout and explain how it works.” Sadowski performed the simulations presented in the paper, while Baldi derived the theory. “The paper also uncovered interesting connections between dropout and other areas of statistics.”

Sadowski, who is now an assistant professor of information and computer science at the University of Hawaii, will be at the 2019 International Joint Conference on Artificial Intelligence (IJCAI 2019) in Macau, China in August to attend the award ceremony.