Nested three-fold cross-validation procedure of Percolator. Each of the three cross-validation training sets are further divided into three nested cross-validation sets, intended to select the most suitable hyperparameters for the SVM training procedure. Just as for the global cross-validation scheme, different permutations of two subsets for training and one subset for testing are used to evaluate each set of hyperparameters.
Granholm et al. BMC Bioinformatics 2012 13(Suppl 16):S3 doi:10.1186/1471-2105-13-S16-S3