Position: PhD Student
Current Institution: Carnegie Mellon University
Abstract: Foundations of Data-Driven Algorithm Design
My research spans artificial intelligence, algorithm design, and computational economics. I am particularly interested in developing machine learning tools for automated algorithm design.
Practitioners often use machine learning techniques to optimize over families of parametrized algorithms, tuning parameters based on typical problems from their domains. Ideally, this results in algorithms with high performance on future problems. Unfortunately, parameter optimization procedures come with few guarantees. To counteract this trend, I am building a theoretical framework for application-specific algorithm design. My work involves characterizing problem-specific structure that I can exploit to provide computational tools with strong statistical guarantees. This task is challenging because across many domains, a small tweak in parameters can cause a cascade of changes in an algorithm’s behavior. As a result, the algorithm’s performance is a non-convex and volatile function of its parameters.
In a related direction, my research also includes data-driven auction design, a special case of algorithm design with tremendous real-world impact. The rise of the internet has led to worldwide participation in electronic marketplaces, which in turn has generated a deluge of consumer data. In my work, I design algorithms that learn consumers’ valuations from a small amount of data. In addition, I develop tools that utilize this data to design auctions and optimize prices, with revenue maximization as the goal.
Finally, I am passionate about protecting against the societal risks that come with the exploitation of data, such as privacy loss. For example, large internet companies regularly tune prices according to consumer data. Without proper precautions, the resulting prices may leak information about consumers’ purchase histories. In my work, I develop algorithms that utilize data while protecting private information contained therein.
Ellen Vitercik is a PhD student in computer science at Carnegie Mellon University, advised by Nina Balcan and Tuomas Sandholm. Her primary research interests are artificial intelligence, machine learning, theoretical computer science, and computational economics. Her honors include a National Science Foundation Graduate Research Fellowship and a Microsoft Research Women’s Fellowship. She receied a bachelor’s degree in mathematics from Columbia University, graduating summa cum laude and Phi Beta Kappa .