Section: Partnerships and Cooperations
International Research Visitors
From September to December 2019 we hosted Vaishnavi Bhargava, a computer science student from the Birla Institute of Technology and Science in Pilani, who worked on “Automatic Neighborhood Design for Localized Model-interpretation”. Her work aimed at finding a set of metrics and procedures to determine the best parameterization of the method LIME for local post-hoc interpretability of machine learning models. The goal of this effort is to inform users of the parameter values (if any) for which a LIME explanation should be trusted because it can faithfully reproduce the behavior of the black-box it tries to explain.