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.

quadratic lssvm

Submitted by reference

LS-SVM with a normalised inhomogenous quadratic kernel, model selection via scaled conjugate gradient descent, minimising the PRESS criterion.

Performance Estimates:

ADA - BER = 0.187971 : AUC = 0.899026

GINA - BER = 0.057844 : AUC = 0.984841

HIVA - BER = 0.244383 : AUC = 0.984024

NOVA - BER = 0.055013 : AUC = 0.986236

SYLVA - BER = 0.007721 : AUC = 0.999052

Dataset Balanced Error Area Under Curve  
Train Valid Test Train Valid Test
ada 0.1682 0.2143 0.1876 0.9164 0.8742 0.8984 agnostic
gina 0 0.0317 0.0509 1 0.994 0.9885 agnostic
hiva 0.0212 0.2535 0.2733 0.9972 0.7253 0.772 agnostic
nova 0.0004 0.064 0.0548 1 0.9955 0.9891 agnostic
sylva 0.0023 0.0045 0.0078 0.9999 0.999 0.9988 agnostic
Overall 0.0384 0.1136 0.1149 0.9827 0.9176 0.9294 agnostic