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.

trees - balanced

Submitted by Joerg Wichard

An ensemble of CART tress,
selected via cross-validation (~150 rounds),
the validation data was included.
The data sets were balanced, no further
preprocessing was applied.

Dataset Balanced Error Area Under Curve  
Train Valid Test Train Valid Test
ada 0.2557 0.2619 0.2686 0.9075 0.88 0.9021 agnostic
gina 0.0614 0.0443 0.0926 0.9847 0.9929 0.9661 agnostic
hiva 0.1525 0.162 0.3031 0.9247 0.9224 0.7638 agnostic
nova 0.1255 0.134 0.142 0.9639 0.9685 0.9449 agnostic
sylva 0.0249 0.0321 0.0348 0.9995 0.9995 0.9983 agnostic
Overall 0.124 0.1269 0.1682 0.9561 0.9527 0.9151 agnostic

This entry is a complete agnostic learning entry.