Newsworthy research: Novel data analysis methods earn National Science Foundation grant

For the third consecutive year, the National Science Foundation (NSF) directed one of its prestigious Faculty Early Career Development (CAREER) Program awards to a Tepper School of Business faculty member. Wolfgang Gatterbauer, assistant professor of business technologies, was granted a five-year, $550,000 award in spring 2016 for his research proposal to develop novel methods to draw conclusions from uncertain and inconsistent data, as part of a cross-campus project titled “Scaling approximate inference and approximation-aware learning.”

Fatma Kilinç-Karzan, associate professor of operations research, and Mustafa Akan, associate professor of operations management, earned the 2015 and 2014 CAREER awards, respectively. “I was truly honored to receive this award from the NSF, and I am looking forward to working on this project with the excellent students and faculty from the Tepper School, School of Computer Science and other areas of campus,” Gatterbauer said.

In recent years, science has witnessed tremendous progress in areas such as knowledge aggregation, information extraction, question-answering systems, computer vision and machine intelligence. These areas rely on methods that make sense of large amounts of uncertain information. However, with expanding size of data, it becomes increasingly difficult to draw meaningful conclusions, and most methods use sampling-based approaches to achieve probabilistic models. The intent of this project is to develop methods that avoid sampling and instead make use of novel algebraic methods that enable existing relational databases to perform approximate probabilistic inference without major adaptations.

Thus, the ultimate goal is to allow practitioners to use and repurpose existing relational database infrastructure to manage uncertain data, rather than requiring dedicated systems. Gatterbauer and his collaborators showed the viability of this approach in a recent series of papers that propose methods, combining theory from linear and relational algebra, that allow existing relational databases to perform approximate probabilistic inference for special cases much more expeditiously than previously.

“It speaks highly of our school that the NSF has repeatedly recognized our faculty with these awards over the years,” Gatterbauer said. “This award is an important funding source that allows young, motivated faculty to focus on foundational scientific problems whose solutions can have far-reaching effects and will ultimately improve business and our everyday lives.”