PASCAL Agnostic Learning
vs.
Prior Knowledge
IJCNN07

The challenge is now over. But it remains open for post-challenge submissions!


IMPORTANT: Entries made since February 1st 2007 might be using validation data, now available for training.

serate rbf lssvm

Submitted by reference

LS-SVM with a radial basis function kernel, model selection via scaled conjugate gradient descent, minimising a smoothed error rate criterion.

Performance Estimates:

ADA - BER = 0.188753 : AUC = 0.899503

GINA - BER = 0.057037 : AUC = 0.985427

HIVA - BER = 0.264468 : AUC = 0.787690

NOVA - BER = 0.064490 : AUC = 0.982184

SYLVA - BER = 0.007858 : AUC = 0.999040

Dataset Balanced Error Area Under Curve  
Train Valid Test Train Valid Test
ada 0.1646 0.2127 0.1882 0.922 0.8742 0.9005 agnostic
gina 0 0.0442 0.0498 1 0.9952 0.9889 agnostic
hiva 0.0187 0.2973 0.2778 0.9988 0.7143 0.7624 agnostic
nova 0 0.048 0.0642 1 0.993 0.9842 agnostic
sylva 0.0022 0.0049 0.0084 0.9999 0.9991 0.9988 agnostic
Overall 0.0371 0.1214 0.1177 0.9841 0.9152 0.927 agnostic