Manifold: A Model-Agnostic Framework for Interpretation and Diagnosis of Machine Learning Models | IEEE Journals & Magazine | IEEE Xplore