Optimizing the efficiency of cross-validation in linear discriminant analysis through selective use of the Sherman-Morrison-Woodbury inversion formula

Cross-validation (CV) is a necessary stage in the development of multivariate discriminant models, but is potentially very time-consuming. Significant time saving is possible by employing update formula to avoid unnecessary recalculations. We show that using the Sherman-Morrison-Woodbury (SMW) inversion formula can sometimes provide additional speed gains. The potential gain depends on the structure of the dataset and CV approach. We recommend comparing rival schemes before starting long computational tasks. Datasets and Matlab® m-files are available at www.metabolomics-nrp.org.uk/ publications.html