5 minutes with Lisa Hellerstein
“It’s great to have powerful computers, but unless you have good algorithms, the power isn’t useful.” As part of this month’s Women in Data Science series, we catch up with Lisa Hellerstein, Professor of Computer Science and Engineering
Lisa Hellerstein is a Professor of Computer Science and Engineering at NYU and an affiliated faculty member at the Center for Data Science. After earning her doctorate at the University of California Berkeley, Hellerstein now specializes in machine learning, algorithms, discrete mathematics, and more.
1. Much of your recent research — and the courses that you teach at NYU — have been focused on algorithm design. How did you discover this field, and why do you like it?
Algorithms are a fundamental part of computer science. I first learned about algorithm design in my undergraduate courses. It’s great to have powerful computers, but unless you have good algorithms, the power isn’t useful. In algorithm design, you are solving puzzles. Sometimes, when you begin to look at a computational problem, you have no idea whether it is even possible to solve that problem with an efficient algorithm. You may go back and forth, thinking one day that it’s possible, and another that it is not. It’s especially satisfying when you gain insight into a problem, and through that insight, are able to develop an efficient algorithm.
2. In addition to your research, you’ve also run a NSF-funded project, “Robust Uncertain Data Management,” for several years. What are your research goals? How does your work here impact different industry areas?
I have a long-standing interest in algorithms for choosing sequences of questions or tests. Think of the 20-questions game: Which question should you ask first? Which should you ask next? The NSF project on data management was a joint project with a researcher in databases at the University of Maryland, Amol Deshpande. Database design also involves sequences of questions. For example, suppose you ask the database to provide you with the records of all people in your organization who make more than $60K per year and who live in New York City. Given a person’s record, should you first check whether the person lives in New York City? Or whether the person makes above $60K per year? How you make that decision can affect how quickly the database is able to provide the requested records. As a theoretician, I work on abstract versions of question and test ordering problems. I aim to develop efficient algorithms for determining optimal or near-optimal orderings, in worst-case, probabilistic, and game theoretic settings. Although my work is theoretical, the problems I address are connected to practical problems in many different areas, including machine learning, databases, and medical diagnosis.