Table 3.

AUC and elapsed time in the Fischer data set.

						95% confidence interval
Method	Platform	Restarts	Elapsed time	AUC	standard error	lower	upper
SHEBA	CPU	-	03 m 47 s	0.877	0.016	0.845	0.909
DaliLite	CPU	-	113 m 34 s	0.876	0.016	0.845	0.908
LOCK2	CPU	-	32 m 12 s	0.874	0.016	0.842	0.907
IR Tableau	CPU	-	< 1 s	0.859	0.015	0.830	0.888
QP Tableau Search	CPU	-	52 m 10 s	0.848	0.018	0.813	0.882
SA Tableau Search	CPU	4096	12 m 07 s	0.837	0.018	0.801	0.872
SA Tableau Search	Tesla C1060	4096	00 m 55 s	0.836	0.018	0.800	0.871
SA Tableau Search	GTX 285	4096	00 m 40 s	0.829	0.018	0.792	0.865
SA Tableau Search	CPU	128	00 m 25 s	0.822	0.019	0.786	0.859
SA Tableau Search	GTX 285	128	00 m 02 s	0.809	0.019	0.771	0.846
SA Tableau Search	Tesla C1060	128	00 m 02 s	0.804	0.019	0.767	0.842
SARF2	CPU	-	19 m 34 s	0.797	0.020	0.759	0.835
YAKUSA	CPU	-	00 m 02 s	0.741	0.021	0.700	0.782
TOPS	CPU	-	01 m 05 s	0.692	0.022	0.649	0.734
TableauSearch	CPU	-	< 1 s	0.662	0.022	0.619	0.705
VAST	CPU	-	02 m 33 s	0.600	0.022	0.557	0.643

Area under the ROC curve (AUC) and elapsed time for different methods for all-against-all comparisons in the Fischer data set. The table is sorted by AUC descending. In the Platform column, either a GPU card is described, or CPU, which is a single core of an AMD Quad Core Opteron (2.3 GHz, 32 GB RAM) running Linux.