Skip to main content

Table 2 Comparison of Core Hunter 2 and 3

From: Core Hunter 3: flexible core subset selection

 

E-NE

DMIN

Time (s)

Coconut

CH2

0.552 ± 3.53e-2

0.501 ± 9.76e-2

27.6 ± 06.0

CH3

0.576 ± 9.35e-5

0.540 ± 0.00e-0

37.5 ± 07.9

CH2L

0.569 ± 5.91e-4

0.548 ± 0.00e-0

31.0 ± 00.1

Maize

CH2

0.416 ± 1.52e-2

0.396 ± 2.46e-2

78.3 ± 10.6

CH3

0.435 ± 2.70e-4

0.409 ± 3.05e-3

74.3 ± 26.5

CH2L

0.429 ± 5.00e-4

0.415 ± 1.11e-3

78.6 ± 02.0

Pea

CH2

0.219 ± 1.49e-3

0.000 ± 0.00e-0

85.6 ± 04.5

CH3

0.338 ± 1.04e-3

0.287 ± 1.34e-2

154.1 ± 49.7

CH2L

0.325 ± 8.21e-4

0.297 ± 0.00e-0

802.3 ± 00.8

  1. CH2 maximizes a weighted index including average and minimum pairwise distance, with equal weight, while CH3 maximizes E-NE. Mean E-NE, DMIN, runtime and corresponding standard deviations are reported for 10 independent executions. The highest obtained E-NE and DMIN value per dataset is shown in bold. CH3 was terminated when no improvements were found during 10 s. For CH2, two alternatives were considered: (a) the same stop condition as for CH3 (CH2); and (b) an absolute runtime limit that was empirically determined per dataset to ensure that the LR replica of MixRep terminated in each run (CH2L)