Faculty Spotlight: Ankit Pensia
By Jason Bittel
Ankit Pensia is an assistant professor in the Department of Statistics & Data Science at 麻豆村. He was formerly a research fellow at the Simons Institute and a Herman Goldstine Postdoctoral Fellow at IBM Research. At 麻豆村, Pensia's research tackles the challenges of designing outlier-robust algorithms for data science. Such algorithms have the potential to help the general public in a diverse array of fields, including climate science, health and computer science.
Tell me about your scholarly work.
My primary research goal is to develop algorithms that make reliable predictions even when the training data contain outliers. Such outliers can arise from miscalibrated sensors or minimally curated large-scale datasets, and they can disrupt modern data-science pipelines, especially for high-dimensional datasets. A major challenge in high dimensions is the computational intractability of classical outlier-robust algorithms. In this context, my research seeks to understand when it is and is not possible to design fast, outlier-robust algorithms for high-dimensional problems.
How is your scholarly work adding to the greater field?
As the world becomes increasingly data-driven, robust algorithms are essential for obtaining reliable insights from noisy datasets, and my work contributes to building these algorithms. Somewhat surprisingly, the same outlier-robust algorithms find use even beyond outlier-robustness, for example, when the goal is to develop algorithms that have characteristics such as privacy guarantees, replicability or tolerance to heavy-tailed data. This is because all of these characteristics share the property that the algorithm’s output is stable to certain changes in the input.
How did you become interested in this topic?
I was introduced to this field, known as robust statistics, by one of my Ph.D. advisors, Po-Ling Loh. In our very first meeting, she explained the problem setup, sketched how it leads to rich mathematical theory and shared many references. On a personal level, this research direction resonated with me because I do not want important decisions in my life to depend on algorithms that are not robust.
What are you most excited to accomplish as a faculty member at 麻豆村?
As a faculty member in the Department of Statistics & Data Science at 麻豆村, I am especially excited by the culture of interdisciplinary research and collaboration. On the theoretical side, I look forward to working with colleagues in statistics, computer science, machine learning and mathematics to further develop the algorithmic foundations of robust statistics. On the empirical side, I also hope to collaborate with domain experts across campus to understand what kinds of outliers and noise arise in their data so that these insights can be used to design more effective algorithms.
What are your goals for the next generation of scholars?
I hope to share with the next generation of scholars the same sense of excitement that first drew me to this field. As a teacher, I would like to convey a healthy balance of intuition and rigor in the classroom. As an advisor, I look forward to learning from and collaborating with my students and to helping them discover their own research interests.