Using Diversities to Model the Reliability of Two-version Machine Learning Systems | IEEE Journals & Magazine | IEEE Xplore