Meet Assistant Professor Mark Sellke in the Harvard Department of Statistics

After completing his PhD at Stanford University in 2022 and a postdoc position at the Institute of Advanced Study in Princeton and at Amazon, Dr. Mark Sellke has joined the Department of Statistics in July 2023 as an Assistant Professor.  To introduce Prof. Sellke to the department, we conducted the following interview (revised and excerpted) in which Mark shares about his research interests, middle school math competitions, and what he’s looking forward to the most this year.  Welcome Prof. Sellke!

1.  Please introduce yourself (e.g. where you are from, your current work, a hobby that's important to you, and a fun fact about you).

I grew up in the Midwest, attended MIT as an undergraduate, and completed my PhD at Stanford University.  One of my main research interests is spin glasses, which is a high dimensional [high dimensional data occurs when the number of features in a data set is higher than the number of observations] probability topic that is also closely related to questions in statistical inference.  The area of spin glasses has its roots in statistical physics. The initial motivation was to study and develop models for certain magnetic materials with unusual properties in 3-dimensional space. A fruitful avenue for making progress has been to treat the objects as high dimensional (as opposed to 3-dimensional), which makes them relevant to modern questions in machine learning and statistics. In my research, I'm also interested in other types of machine learning theory, including multi-armed bandits.

For hobbies, I enjoy racket sports. I used to play tennis a lot, and I recently started playing squash. Another favorite hobby of mine is chess, which I hear is popular around the department!

A fun fact about me is that I was once featured on the social media page of a local food truck in Cambridge. They took a photo of me when I was waiting in their line on a below-freezing winter day, wearing just a t-shirt, shorts, and flip-flops (I wasn’t very sensitive to cold when I was younger).

2.  Describe some experiences (or mentors) that sparked your interest in statistics.

I've enjoyed probability since middle school because the problems are a lot of fun and there are many interesting paradoxes, such as Simpson’s paradox and the Monty Hall problem.  In these problems, you distill a complicated situation into a single number to gain some understanding about it.  While in middle school, I also had a lot of fun participating in math competitions, starting with MATHCOUNTS, which was the primary math competition for middle school kids in the US.

One formative experience after college was when I spent a summer at Microsoft Research just before starting my PhD.  During this time, I explored several interesting problems in statistics and machine learning, including problems on high dimensional optimization and real-time decision making.  While my research has evolved since this experience, I still enjoy thinking about some of these themes in my research.

3.  Describe your current research and/or teaching interests. You can discuss a particular project, if you'd like.

To simply describe my work, I am trying to understand high dimensional optimization from the average case perspective rather than the worst-case perspective.  Traditionally in optimization theory, you are trying to optimize a high dimensional function so that the algorithm works in all possible cases. From this perspective, our algorithm might have to bend over backwards to do well for an unrealistic worst-case scenario, which we can avoid by studying the average case. I am in the process of developing a general theory that identifies the best stable algorithms (meaning that if you change the input, the output of the algorithm doesn’t fluctuate very much) for a wide class of such optimization problems.  This spring I'm looking forward to teaching a class with a broad focus on average-case optimization. I plan to survey interesting qualitative behaviors that can appear in these types of problems.

4.  What made you decide to join Harvard Stats? What are you looking forward to the most for this year?

I am excited to interact and collaborate with the brilliant students and colleagues in the department and campus-wide. Everyone has been very friendly and welcoming!