Table 1.

Summary of classification results. Results of multi-class SVM: the empirical training loss <L_tr> averaged over all pairwise classifiers, the average empirical test loss <L_tst>, the average number of support vectors <N_sv>, and the winning model (with the highest number of votes from all pairwise classifiers). For the definition of robustness see Methods.

	E. coli	C. elegans	S. cerevisiae
L tr	1.6%	0.5%	2.1%
L tst	1.6%	0.5%	1.8%
N sv	109	51	106
Winner	Kumar [13]	MZ [14]	Sole [15]
Robustness	1.0	0.97	0.64