October 21, 2016
Research in complex computational problems snares Packard honors for UW’s Thomas Rothvoss
The David and Lucile Packard Foundation has awarded a prestigious fellowship to University of Washington assistant professor Thomas Rothvoss to fuel his passion to balance precision and efficiency in complex mathematical calculations. The Packard Foundation Fellowships for Science and Engineering honor early-career academics pursuing innovative research in all fields of science and engineering.
“It’s a great honor — and frankly, I’m still digesting the news,” said Rothvoss, who has joint appointments with the Department of Mathematics and the Department of Computer Science & Engineering, or CSE.
Rothvoss’ research probes the theoretical limits of computer algorithms designed to analyze large, complex datasets. He wants to understand when these algorithms operate quickly and efficiently and when they cannot. In situations where efficiency craters, he develops alternative mathematical methods that can approximate an optimal answer.
“This is really a problem of optimization,” said Rothvoss. “Let’s say you have a lot of objects of different weights and sizes — and boxes to pack them in. You want to optimize the packing process, grouping the objects into the smallest number of boxes as possible as quickly as possible.”
It could be possible to write an algorithm to perform this calculation and produce a plan for packing the objects in boxes, Rothvoss said. But, the scenario is so complex that the algorithm would be sluggish and inefficient.
“A better approach is to recognize these limitations and come up with algorithms that approximate an ideal answer,” said Rothvoss. “Find an efficient algorithm that gets you close to that optimum. That’s my specialty.”
Efficiency problems plague the era of big data. For example, today’s best DNA sequencers produce trillions of short sequences of DNA letters. It falls to computer scientists to write algorithms that can query these giant libraries of letters and assemble complete genome sequences of an individual or organism.
“This field is called ‘combinatorial optimization,'” said Rothvoss. “And it lies at the intersection of theoretical computer science and mathematics.”
A German native, Rothvoss became interested in the theoretical limits of computer science as an undergraduate at the Technical University of Dortmund. That meant spending more time in mathematics courses than the average computer science student, and he quickly became “hooked” on approximation methods.
Rothvoss earned his doctorate in 2009 from the Swiss Federal Institute of Technology at Lausanne, and later worked as a postdoctoral researcher both there and at the Massachusetts Institute of Technology. He came to the UW Department of Mathematics in 2014 as an assistant professor, and received a joint appointment with CSE in 2015.
As a Packard fellow, Rothvoss will receive $875,000 over five years. The Packard Foundation selects just 18 fellows each year from an initial pool of 100 nominees — two each from 50 universities. A 12-member panel of scientists and engineers selects the winners.
The fellows program began in 1988 to give early-career researchers — that is, those within the first three years of their first academic appointment — flexibility to pursue “risky” research avenues. As research funds atrophy from traditional sources, these fellowships become increasingly important for unlocking innovation.
“Basically, this means for the next five years that I can spend less time writing grants and more time doing research with my students,” said Rothvoss. “And that is wonderful news.”
Rothvoss is the UW’s 10th Packard fellow, eight of whom are still affiliated with the university. He is the first from the UW Department of Mathematics, but the third for CSE. Past Packard fellows from CSE are Rajesh Rao and Christopher Diorio, who is now CEO of Impinj, a UW spinoff.
###
For more information, contact Rothvoss at 206-543-1723 or rothvoss@uw.edu.
Tag(s): awards • College of Arts & Sciences • College of Engineering • Department of Mathematics • Paul G. Allen School of Computer Science & Engineering • Thomas Rothvoss