The Dutch Optimization Seminar is an initiative to bring together researchers from the Netherlands and beyond, with topics focusing on optimization in a broad sense. We would like to invite all researchers, especially PhD students, who are working on related topics to join the events. We hereby announce the following lecture, given by Ilker Birbil (University of Amsterdam):
Speaker: Ilker Birbil (University of Amsterdam)
Title: Counterfactual explanations using optimization with constraint learning
Date: Thursday August 254:00 p.m. CET
Counterfactual explanations embody one of many interpretability techniques that are receiving increasing attention from the machine learning community. Their potential to make model predictions more sensitive to the user is considered invaluable. In order to increase their adoption in practice, several criteria to which counterfactual explanations should conform have been put forward in the literature. We propose counterfactual explanations using constraint learning optimization, a generic and flexible approach that meets all these criteria and leaves room for further extensions. Specifically, we discuss how we can take advantage of an optimization with a constraint learning framework for the generation of counterfactual explanations, and how the components of this framework easily match the criteria. We also propose new and innovative modeling approaches to deal with proximity and diversity of data varieties, which are two key criteria for practical counterfactual explanations. We test our approaches on several datasets and present our results in a case study. Compared to a current state-of-the-art method, our modeling approach showed superior overall performance in terms of several evaluation metrics proposed in related works while allowing more room for flexibility.