ABSTRACT:

Machine learning has been increasingly deployed for high-stakes decisions that deeply impact people’s lives. My research focuses on developing interpretable algorithms and pipelines to ensure the safe and efficient utilization of machine learning models in the decision-making process. In this talk, I will introduce a new paradigm, called learning the Rashomon set, which finds and stores all models within epsilon of the optimal loss. I will present algorithms for finding optimal models and Rashomon sets, discuss how this new paradigm can break the interaction bottleneck between users and ML algorithms, and provide examples of its applications.

ZOOM LINK (if unable to attend in-person): https://upenn.zoom.us/j/99641494762