Fast algorithms for robust estimation with large samples of multivariate observations. Estimation of the geometric median, robust k-Gmedian clustering, and robust PCA based on the Gmedian covariation matrix.
Changes to version 1.2.3
Changes to version 1.2.2
Changes to version 1.2.1
Add robust estimation of the explained variance (standard deviation) for each principal axis in functions ‘WeiszfeldCov’ and ‘GmedianCov’. The robust estimation is obtained thanks to function robustbase::scaleTau2
Add the possibility to get a non negative estimated covariation matrix in function ‘GmedianCov’ by considering an approximation to the projection on the convex closed cone of the non negative matrices at each iteration of the Robbins-Monro algorithm.
Changes to Version 1.2
Add function 'Weiszfeld' based on Weiszfeld's algorithm for computing the geometric median. Slower than 'Gmedian' but more accurate for small and moderate sample sizes.
Add function 'WeiszfeldCov' based on Weiszfeld's algorithm for computing the median covariation matrix. Slower than 'GmedianCov' but more accurate for small and moderate sample sizes. Not appropriate for high dimension (d>500) data.
Changes to Version 1.1
Fixed bug in C++ related to the (forbidden) use of function 'sqrt' for integers
Use function 'eig_sym' from package Rspectra instead of package rARPACK