Figure 3.

The algorithm for cross-validation. Given a set, K, of tuples (<a onClick="popup('http://www.biomedcentral.com/1471-2105/13/S16/S3/mathml/M1','MathML',630,470);return false;" target="_blank" href="http://www.biomedcentral.com/1471-2105/13/S16/S3/mathml/M1">View MathML</a>, isDecoy, g) representing PSMs, where <a onClick="popup('http://www.biomedcentral.com/1471-2105/13/S16/S3/mathml/M1','MathML',630,470);return false;" target="_blank" href="http://www.biomedcentral.com/1471-2105/13/S16/S3/mathml/M1">View MathML</a> is a vector of PSM features, isDecoy is a boolean variable indicative of whether the PSM is a decoy PSM or not, and g ∈ {1, 2, 3} is a tag indicating which cross-validation set the tuple should be allocated to, the algorithm returns a set of PSMs. The function InternalCrossValidation() is used for nested cross-validation within the training set and returns the most efficient set of learning hyperparameters. The SVMTrain() function uses the training set and hyperparameters and returns the learned feature weights needed to score the PSM.

Granholm et al. BMC Bioinformatics 2012 13(Suppl 16):S3   doi:10.1186/1471-2105-13-S16-S3