Feature Selection Challenge

The challenge is over, it ended on December 12, 2003, but the web site is now open again for people who want to benchmark their system against the challenge entries.

BayesNN-DFT-combo

Submitted by Radford Neal & Jianguo Zhang

Combination of Bayesian neural networks and classification based on
Bayesian clustering with a Dirichlet diffusion tree model. A
Dirichlet diffusion tree method is used for Arcene. Bayesian neural
networks (as in BayesNN-large) are used for Gisette, Dexter, and
Dorothea. For Madelon, the class probabilities from a Bayesian neural
network and from a Dirichlet diffusion tree method are averaged, then
thresholded to produce predictions.

Dataset	Balanced Error			Area Under Curve			Features
	Train	Valid	Test	Train	Valid	Test	#	%
arcene	0.0000	0.0722	0.1330	1.0000	0.9769	0.9348	10000	100.00
gisette	0.0007	0.0160	0.0129	1.0000	0.9978	0.9990	5000	100.00
dexter	0.0000	0.0533	0.0390	1.0000	0.9846	0.9901	303	1.52
dorothea	0.0187	0.0547	0.0854	0.9989	0.9736	0.9592	100000	100.00
madelon	0.0000	0.0667	0.0717	1.0000	0.9804	0.9782	500	100.00
overall	0.0039	0.0526	0.0684	0.9998	0.9827	0.9722		80.30