Jelena Diakonikolas

Jelena Diakonikolas wants to strengthen the algorithms that power machine learning models. She recently received an NSF CAREER Award to develop new algorithms that perform well under uncertain and changing circumstances.

Her CAREER award was inspired by a Data Science Institute symposium on AI for agriculture and forestry, where researchers voiced challenges with building models that are robust in ambiguous and changing data environments, echoing broader issues with how machine learning models are trained and used today. Many algorithms assume that training data and real-world data are the same, which often isn’t true. Additionally, data may change once a model is used. For instance, implementing a new policy based on collected data may change the data itself. Diakonikolas’s work will address limitations in current machine learning approaches, creating algorithms that can power models that are accurate and stable.

Growing up in Serbia, Diakonikolas excelled at math and computer sciences, but lacked exposure and opportunities for rigorous academic research. Her decision to move to the U.S. for graduate school helped her grow as a researcher and build confidence. As a teacher, she encourages her students to address uncertainty in their lives by keeping an open mind and trying new things.

“Some students fear the unknown, or they are just really nervous about what classes to take. I can remember that uncertainty. At that point in your life, you think every decision you make determines your entire future. I try to explain to students that not knowing something is okay. The important thing is to keep learning.”