Sarah Dean: Understanding the fundamentals of data-driven methods for control and decision-making

Assistant professor, Computer Science

Academic focus: Machine Learning, Dynamical Systems, and Control 

Research summary: I study the interplay between optimization, machine learning, and dynamics in real-world systems. My research focuses on understanding the fundamentals of data-driven methods for control and decision-making, inspired by applications ranging from robotics to recommendation systems. One thrust of my work involves investigating how to guarantee safety and robustness in control and reinforcement learning. This is motivated by the desire to incorporate the successes of machine learning into safety-critical settings. Another thrust includes exploring how to ensure that human values like fairness and agency are present in dynamical and data-driven systems. This is motivated by the impacts of algorithms in our digital world.

What inspired you to pursue a career in this field? Math was always one of my favorite subjects -- first in elementary school because it came easily to me, and later in college because, when it was no longer easy, I realized how beautiful and intellectually rich it is. At the same time, I grew up in a family of engineers which oriented me towards building things and solving "real" problems. This combination led me first towards electrical engineering and later to dynamical systems and machine learning.

As to what inspired the particular path my career has taken, I largely credit the people I've encountered along the way. While I was intentional about getting involved in research early on, I never had a master plan. Instead, I pursued whatever opportunities were most interesting -- from high school research on dragonfly perception to undergraduate work on sensors for robot locomotion and mountain lion tracking. During my PhD I was lucky to continue to explore, with collaborations in computational microscopy, autonomous car racing, and ethical recommendation systems. What began as a casual interest in the impact of technology on society led to founding a cross-disciplinary student group and a shift in my research agenda. I hope as a professor I can continue following my curiosity and find inspiration from my colleagues here at Cornell.

What are you most looking forward to as a Cornell Engineering faculty member? I am excited (and nervous!) about teaching and mentoring students. I really enjoy the process of teaching -- I like how it forces you to deeply understand subject material. The best part comes from interacting with students, who will challenge you to consider things from different angles. Ultimately, I love teaching because I love learning. 

I am particularly excited about teaching and mentoring students in machine learning. In some ways, it's an old field, with many core ideas present by the 1950s-60s. In other ways, it's a very new field, with an explosion of progress and interest in the past couple of decades. I think we are still figuring out what the core ideas should be and how to teach them. There are several challenges: machine learning relies largely on continuous mathematics whereas the CS curriculum emphasizes discrete math. The hype surrounding "big data" and "AI" technologies obscures their limitations, which calls for teaching students how to take a critical perspective. I'm looking forward to facing these challenges head on when I teach Introduction to Reinforcement Learning this spring.

What do you like to do when you’re not working? I spend a lot of time reading: long form journalism, essays, nonfiction, and speculative fiction. Despite my lack of athletic talent, I like to be active and outside: hiking, yoga, biking, sailing, ice skating, and cross-country skiing. I am especially excited for winter sports now that I'm back living somewhere it snows!

Learn more about Sarah's research at her faculty page.