Recovering valid clusters with ISODATA supervised by the CAIC
Charles S. Carman, Michael Merickel · Proceedings of the Annual International Conference of the IEEE Engineering in Medicine and Biology Society · 1988
The authors developed an unsupervised clustering method that is a variant of the well known ISODATA clustering algorithm. They replace the heuristic rules that control ISODATA with rules that search for the minimum value of an information theoretic criterion. The criterion investigated in this study is the Consistent Akaike's Information Criterion (CAIC). The CAIC is a measure of the global fit of a cluster model to the input data, and the smallest CAIC value suggests the best fit. The authors tested the method on both multivariate Gaussian and real-world data, including MR (magnetic resonance) images of aortas in vivo.>