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.

linear lssvm

Submitted by reference

LS-SVM with a normalised linear kernel, model selectio via scaled conjugate gradient descent, minimising the PRESS criterion.

Performance estimates:

ADA - BER = 0.193458 : AUC = 0.889171

GINA - BER = 0.132410 : AUC = 0.936352

HIVA - BER = 0.254677 : AUC = 0.807146

NOVA - BER = 0.049135 : AUC = 0.987761

SYLVA - BER = 0.014933 : AUC = 0.998180

Dataset Balanced Error Area Under Curve  
Train Valid Test Train Valid Test
ada 0.1916 0.2221 0.1966 0.89 0.8639 0.8897 agnostic
gina 0.1014 0.1273 0.126 0.9609 0.9461 0.9461 agnostic
hiva 0.0981 0.3311 0.2832 0.9648 0.699 0.7598 agnostic
nova 0.0004 0.044 0.0481 1 0.9968 0.9894 agnostic
sylva 0.0116 0.0069 0.0161 0.9987 0.998 0.9978 agnostic
Overall 0.0806 0.1463 0.134 0.9629 0.9008 0.9166 agnostic