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.

Data Grid (CMA)

Submitted by Marc B oulle

Data Grid (CMA)

Prior submission (prior for ada, gina and sylva, agnostic for hiva and nova)

Data Grids extend the MODL discretization and value grouping methods to the multivariate case (paper submitted to IJCNN 2007).
An ensemble of data grid is averaged according to the compression-based averaging schema (CMA) (see SNB(CMA) submission)

Preprocessing for sylva: the 80 soil_type binary variables are merged into two categorical variables


BER guess:
ada: 0.192
gina: 0.140
hiva: 0.310 (agnostic)
nova: 0.135 (agnostic)
sylva: 0.008

Dataset Balanced Error Area Under Curve  
Train Valid Test Train Valid Test
ada 0.1734 0.1994 0.1756 0.8507 0.8045 0.8464 prior
gina 0.1052 0.089 0.1254 0.9588 0.972 0.9479 prior
hiva 0.2651 0.3276 0.3242 0.7659 0.7647 0.717 agnostic
nova 0.1094 0.096 0.1229 0.9226 0.9338 0.9159 agnostic
sylva 0.0205 0.009 0.0228 0.981 0.9838 0.9798 prior
Overall 0.1347 0.1442 0.1542 0.8958 0.8917 0.8814 prior

This entry is a complete prior knowledge entry.