Dataset Corel African

Basic characteristics Corel African

7947

instances

2000

bags

9

features

100

positive bags

between 2 and 13

instances per bag

1900

negative bags

Results Corel African

AUC dataset 110
classifier error
AUC EER meanPrec avCost
APR 50.5 (0.0) 49.5 (0.0) 5.1 (0.0) 24.9 (0.0)
DiverseDensity (10.000000) 85.6 (0.1) 21.2 (0.6) 43.8 (0.9) 11.5 (0.4)
EMDD (10.000000) 91.5 (0.4) 16.1 (1.2) 52.7 (1.3) 8.4 (0.6)
MILBoost (100 rounds) 88.9 (0.6) 18.4 (1.4) 48.4 (3.1) 9.6 (0.6)
Citation 1NN, c=1 80.4 (1.6) 19.9 (1.6) 32.2 (3.0) 14.3 (0.9)
Citation 3NN, c=5 86.3 (0.6) 17.9 (0.8) 49.2 (0.7) 11.3 (0.4)
MI-SVM p=1.000000 63.4 (2.0) 40.2 (2.6) 11.7 (1.2) 19.6 (0.5)
MI-SVM r=10.000000 88.1 (1.1) 18.7 (1.5) 37.6 (0.8) 10.2 (0.6)
MILES p=1.000000 88.6 (0.5) 18.5 (0.9) 48.0 (1.2) 9.8 (0.3)
MILES r=10.000000 58.9 (10.0) 43.7 (9.6) 8.9 (3.2) 21.4 (2.1)
SVmil mahal.K 56.7 (0.9) 43.1 (1.7) 10.9 (1.1) 21.7 (0.3)
SVmil haussd.K 68.7 (2.1) 35.7 (1.6) 11.5 (0.5) 18.1 (0.3)
SVmil emd.K 78.1 (0.0) 27.9 (0.0) 16.7 (0.0) 15.3 (0.0)
SimpleMIL with Logistic2 75.7 (0.2) 29.6 (0.8) 15.5 (0.5) 16.2 (0.2)
mean-inst+Logistic2 83.2 (0.3) 23.6 (1.2) 25.8 (0.9) 13.0 (0.1)
extremes+Logistic2 90.0 (0.2) 18.4 (0.8) 49.1 (0.5) 9.5 (0.1)
cov-coef+Logistic2 86.5 (0.9) 21.5 (2.6) 35.3 (2.1) 11.2 (0.3)
Bag of Words (k=10)+Logistic2 76.0 (2.4) 30.8 (2.2) 21.7 (3.7) 15.8 (1.0)
Bag of Words (k=100)+Logistic2 83.4 (2.3) 24.2 (3.8) 40.4 (3.0) 12.4 (1.2)
PposteriorMIL 75.7 (2.7) 31.0 (2.0) 25.4 (0.8) 16.0 (0.9)
Dissim.repr. AUC dataset 110
classifier error
AUC EER meanPrec avCost
minmin.K+Logistic2 95.7 (0.4) 11.2 (0.6) 74.6 (1.7) 5.2 (0.1)
summin.K+Logistic2 91.0 (0.8) 11.3 (1.0) 60.7 (1.3) 7.7 (0.5)
meanmin.K+Logistic2 91.5 (1.2) 11.6 (0.5) 58.2 (1.3) 7.6 (0.7)
meanmean.K+Logistic2 91.5 (0.2) 16.2 (0.6) 63.2 (0.7) 8.3 (0.3)
mahal.K+Logistic2 51.0 (2.0) 49.2 (1.3) 7.7 (0.4) 23.2 (0.4)
haussd.K+Logistic2 71.0 (2.3) 32.4 (2.3) 21.6 (4.7) 17.2 (0.9)
emd.K+Logistic2 82.0 (2.6) 19.5 (2.6) 40.4 (2.3) 13.4 (1.4)
linass.K+Logistic2 - - - -