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.

cubic lssvm

Submitted by reference

LS-SVM with a normalised inhomogenous cubic kernel. Model selection using scaled conjugate gradient descent, minimising the PRESS criterion.

Performance Estimates:

ADA - BER = 0.188008 : AUC = 0.899261

GINA - BER = 0.053219 : AUC = 0.986967

HIVA - BER = 0.252270 : AUC = 0.805144

NOVA - BER = 0.056941 : AUC = 0.985444

SYLVA - BER = 0.007820 : AUC = 0.999049

Dataset Balanced Error Area Under Curve  
Train Valid Test Train Valid Test
ada 0.1644 0.2046 0.1889 0.922 0.8726 0.8991 agnostic
gina 0 0.0285 0.0466 1 0.9955 0.99 agnostic
hiva 0.0158 0.2467 0.273 0.999 0.7486 0.7742 agnostic
nova 0.0004 0.044 0.0562 1 0.9947 0.9881 agnostic
sylva 0.0021 0.0045 0.0077 0.9999 0.9991 0.9988 agnostic
Overall 0.0365 0.1057 0.1145 0.9842 0.9221 0.9301 agnostic