Feature Attribution with Necessity and Sufficiency via Dual-stage Perturbation Test for Causal Explanation

arXiv:2402.08845v1 Announce Type: new
Abstract: We investigate the problem of explainability in machine learning.To address this problem, Feature Attribution Methods (FAMs) measure the contribution of each feature through a perturbation test, where the difference in prediction is compared under different perturbations.However, such perturbation tests may not accurately distinguish the contributions of different features, when their change in prediction is the same after perturbation.In order to enhance the ability of FAMs to distinguish different features' contributions in this challenging setting, we …

abstract arxiv attribution cs.lg difference explainability feature machine machine learning prediction stage stat.me test tests through type via