Erik Sudderth Wins NSF CAREER Award
- Posted by Jesse Polhemus
- on April 1, 2014
Assistant Professor Erik Sudderth of Brown University’s Computer Science Department has just won a National Science Foundation CAREER Award for his work on Bayesian nonparametric learning for large-scale structure discovery. It’s accompanied by a grant in the expected amount of more than $509,000. He joins multiple previous Brown CS faculty winners, including (most recently) James Hays, Ben Raphael, and Chad Jenkins. CAREER Awards are the most prestigious awards given by the National Science Foundation (NSF) in support of outstanding junior faculty teacher-scholars who excel at research, education, and integration of the two within the context of an organizational mission.
The motivations for Sudderth’s research start with very large datasets, which could include anything from the videos available on YouTube to the complete corpus of New York Times articles. Parametric statistical learning algorithms work by tuning model parameters to match a user-specified list of properties, or "statistics", of the data. When these algorithms are used to analyze images and video, for instance, humans are required to laboriously collect examples of objects of interest (for example, people, cars, and buildings). “This puts real limits on what can be learned from even very big datasets,” Erik explains, “because the model’s structure has to be manually specified by experts.”
A nonparametric model, however, allows its structure and complexity to be determined from the data itself, so it can grow naturally as the data grows. This allows for algorithms that are capable of “unsupervised” learning, and because less manual supervision is needed, such methods are much more broadly applicable.
The real-world applications for models of this kind are almost limitless: helping computers analyze photographs to differentiate objects from their surroundings, or allowing robots to determine human cognitive states based on facial expressions, or finding communities within social networks by analyzing patterns of collaboration.
“Erik’s innovative research is highly regarded in both computer science and statistics,” comments BrownCS Department Chair Roberto Tamassia. “The prestigious NSF CAREER award is one more indication that Erik is a leader in the important field of Bayesian nonparametric statistical methods.”
If laypeople find the mathematical and computational methods underlying this work a bit daunting, Sudderth already has their needs in mind. “We’re very eager,” he says, “to put useful tools into the hands of people who don’t yet know what nonparametric methods can provide. The five-year term of the grant lets us take a long-term perspective and address the full data analysis process, from models to algorithms to usable software.”
In addition to supporting research, the CAREER grant funds a three-pronged outreach and education plan that includes: (1) an accessible Python software package to allow for easier data analysis, (2) interdisciplinary research projects involving undergraduate students with training in other sciences or the humanities, and (3) two week-long summer schools on Bayesian nonparametrics to be held at Brown's Institute for Computational and Experimental Research in Mathematics (ICERM).
Sudderth’s colleagues are eager to see the project begin. “Erik does excellent work on all aspects of Bayesian nonparametric models,” says Professor Michael Littman, “from devising new mathematical structures, to applying them to interesting problems in text and vision processing, to developing faster algorithms that handle larger and more complex problems, to providing toolkits so others can leverage these advances in their own work. I'm delighted that the NSF recognized his contributions and promise with a prestigious CAREER award.”
“This is a big honor,” Erik concludes. “This award is about making interdisciplinary links. It’s vital for computer scientists to understand how our code and algorithms are challenged by complicated, messy datasets, and it’s equally important for those in other fields to see how computer science can be used to help understand their data. I’m extremely excited.”