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.

PF+GBT

Submitted by Vladimir Martyanov

Gradient Boosted Trees with parameter optimization via Particle Filtering

Dataset Balanced Error Area Under Curve  
Train Valid Test Train Valid Test
ada 0.1704 0.2079 0.1828 0.8285 0.8044 0.8169 agnostic
gina 0 0.0348 0.0471 1 0.9616 0.9527 agnostic
hiva 0.1756 0.3141 0.3035 0.8295 0.6836 0.6978 agnostic
nova 0.0736 0.106 0.1084 0.932 0.8901 0.8892 agnostic
sylva 0.0039 0.0172 0.0108 0.996 0.9826 0.9892 agnostic
Overall 0.0847 0.136 0.1305 0.9172 0.8645 0.8692 agnostic

This entry is a complete agnostic learning entry.