About: Correlative Matrix Mapping (CMM) provides a supervised linear data mapping into a Euclidean subspace of given dimension. Applications include denoising, visualization, label-specific data preprocessing, and assessment of data attribute pairs relevant for the supervised mapping. Solving auto-association problems yields linear multidimensional scaling, similar to PCA, but usually with more faithful low-dimensional mappings.

Tue Jul 5 14:40:03 CEST 2011 - Bugfixes and cleanups

• single precision data affected pinv(). Now fairer using double precision.
• early stopping did not work properly; now fixed
• Hessian update mode globally controlled via hessmode, 'lbfgs' / 'bfgs'
• distmat.m corrected for rounding problems and extended to distmat(X,Y)
• replaced files: corv.m + corvgrad.m -> corvg.m
• removed unused files: corrmat.m, splitdata.m, traforankapply.m