Measurably Stronger Explanation Reliability Via Model Canonization | IEEE Conference Publication | IEEE Xplore