Average training error, test error, and length of four complete signatures as a function of ρ for two algorithms, SPLR (top) and SPLP (bottom). Both errors are expressed as percent 100 × ((FP+FN)/N) where FP is false positive, FN is false negative, and N is the total number of tests. Training error and length results are averages for the four complete signatures. Test errors are averages of four signatures cross-validated three times.