Full documentation is available at: http://www.colbyimaging.com/wiki/statistics/msvm-rfe
This package contains an R implementation of the mSVM-RFE algorithm (Duan et al., 2005), including the option to cut the features by half each round (instead of one-by-one) if there are many features.
The main function is adapted from http://www.uccor.edu.ar/paginas/seminarios/Software/SVM_RFE_R_implementation.pdf.
Also included are tools for wrapping the feature ranking/selection process in an external layer of cross-validation for obtaining unbiased estimates of generalization error/accuracy (See Ambroise et al., 2002).
This entire process can be massively parallelized, and an example implementation is included that uses the SGE cluster interface.
Contact: John Colby