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 linear lssvm

Submitted by reference

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

Performance Estimates:

ADA - BER = 0.195520 : AUC = 0.888985

GINA - BER = 0.131770 : AUC = 0.935318

HIVA - BER = 0.267429 : AUC = 0.795139

NOVA - BER = 0.048343 : AUC = 0.987852

SYLVA - BER = 0.013675 : AUC = 0.998229

Dataset Balanced Error Area Under Curve  
Train Valid Test Train Valid Test
ada 0.1925 0.2172 0.1962 0.8894 0.8644 0.8898 agnostic
gina 0.0825 0.1273 0.123 0.9692 0.94 0.9452 agnostic
hiva 0.0574 0.3566 0.296 0.9895 0.6392 0.7553 agnostic
nova 0.0004 0.044 0.0481 1 0.9968 0.9894 agnostic
sylva 0.0113 0.0123 0.015 0.9987 0.9982 0.9978 agnostic
Overall 0.0688 0.1515 0.1357 0.9693 0.8877 0.9155 agnostic