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.

SVM-RBF

Submitted by Vojtech Franc

SVM, RBF kernel, validation data used for model selection.
SVM is modified such that each class is represented by
an extra classifier like in multi-class case which allows
to use the information from mlabel.

Dataset Balanced Error Area Under Curve  
Train Valid Test Train Valid Test
ada 0.1537 0.1966 0.1915 0.9307 0.8744 0.897 prior
gina 0 0.0031 0.023 1 0.9992 0.9962 prior
hiva 0.1036 0.3027 0.2827 0.9652 0.7952 0.7707 agnostic
nova 0.0014 0.084 0.0877 1 0.9906 0.9834 agnostic
sylva 0.0088 0.018 0.0205 0.998 0.9984 0.9971 agnostic
Overall 0.0535 0.1209 0.1211 0.9788 0.9315 0.9289 prior

This entry is a complete prior knowledge entry.