Carnegie Mellon University
April 08, 2016

Gatterbauer Receives NSF Early Career Award

Five-Year Grant Enables New Conclusions To Be Drawn from Data

The National Science Foundation recently awarded Wolfgang Gatterbauer a Faculty Early Career Development (CAREER) award for his research proposal to develop novel methods to draw conclusions from uncertain and inconsistent data.

Wolfgang Gatterbauer
Wolfgang Gatterbauer

Gatterbauer, an assistant professor of business technologies at the Tepper School of Business and by courtesy the School of Computer Science, received a five-year, $550,000 award for the cross-campus project titled "Scaling approximate inference and approximation-aware learning."

The Faculty Early Career Development (CAREER) Program is a Foundation-wide activity that offers the National Science Foundation's most prestigious awards in support of junior faculty who exemplify the role of teacher-scholars through outstanding research, excellent education and the integration of education and research. Such activities should build a firm foundation for a lifetime of leadership in integrating education and research.

"A CAREER award is a tribute to the quality of Wolfgang's research, and signifies the overall potential of his future research agenda," said Robert Dammon, dean of the Tepper School. "Our school and Carnegie Mellon on the whole remain a nurturing environment for emerging leaders in their fields."

"It is a great honor 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, Computer Science and other areas of campus," said Gatterbauer, who joined the faculty in 2011.

Recent years have witnessed tremendous progress in areas such as information extraction, knowledge aggregation, question-answering systems, computer vision and machine intelligence; however, probabilistic inference remains a key bottleneck and is often performed today with sampling methods. The intent of this project is to develop methods that enable existing relational databases to perform approximate probabilistic inference without any need for sampling.

The ultimate goal is for practitioners to be able to use and re-purpose today's widely existing relational database infrastructure instead of needing dedicated systems to deal with uncertain data. Gatterbauer and his collaborators have already demonstrated the feasibility of this approach in a recent series of papers.

Combining theory from linear and relational algebra, those papers proposed methods that allow existing relational databases to perform approximate probabilistic inference for special cases with considerable speed-up over prior methods. This project aims to generalize these methods and complement them with similarly scalable methods for parameter learning that are "approximation-aware." Thus, instead of treating the learning and the inference steps separately, the goal is to use the approximation methods developed for inference also for model learning, all within the same framework.

Due to its interdisciplinary nature, this project will draw upon the Tepper School, Computer Science and many areas of campus, helping to not only connect and grow the university community but also perhaps result in conclusions with global ramifications.

"Wolfgang comes from the world of computer science," Dammon said. "When we brought him to the Tepper School, we wanted to strengthen our research and our students' understanding of the role business technology plays in modern organizations. We've seen how Wolfgang provides deep, modern insights into that role."

"Wolfgang's work is an excellent bridge between theory and practice: he devises elegant, theoretical solutions, to real, practical problems," said Christos Faloutsos, professor of computer science. "His past and proposed work on linearized 'belief propagation' makes a big difference with respect to execution speed and convergence properties, and it is easily applicable to practical problems such as fraud detection in social networks."

This marks the third such NSF award to Carnegie Mellon faculty regarding data and inference since mid-February. Six weeks ago, the NSF bestowed CAREER awards to two assistant professors in the Department of Statistics of the Dietrich College of Humanities and Social Sciences, Jing Lei and Ryan Tibshirani.

Other current Tepper School faculty members who have won NSF CAREER Awards include Fatma Kılınç-Karzan (2015), assistant professor of operations research, Mustafa Akan (2014), associate professor of operations management, Javier Peña (2005), Bajaj Family Chair in Operations Research, and R. Ravi (1996), Andris A. Zoltners Professor of Business, Rohet Tolani Distinguished Professor and professor of operations research and computer science.