WCCI Performance Prediction Challenge

The challenge is over, but a new challenge is on-going using the same datasets, check it out!

BNN-DR

Submitted by Radford Neal

Bayesian neural network with three hidden layers. The first layer is linear, and reduces dimensionality to ten, the following layers allow non-linear mappings from this ten dimensional space. Some special stuff is done for NOVA.

Dataset Balanced Error Test guess Guess error Test score Area Under Curve
Train Valid Test Train Valid Test
ada 0.146 0.1898 0.1761 0.1648 0.0113 0.1874 0.9321 0.9021 0.9096
gina 0 0.0572 0.0485 0.0598 0.0114 0.0598 1 0.9819 0.9878
hiva 0.0353 0.3768 0.294 0.2683 0.0257 0.3191 0.9931 0.7347 0.7548
nova 0 0.07 0.0557 0.1179 0.0622 0.1179 1 0.9806 0.9861
sylva 0.0041 0.0037 0.0066 0.0072 0.0007 0.0071 0.9999 0.9992 0.9991
Overall 0.0371 0.1395 0.1161 0.1236 0.0223 0.1383 0.985 0.9197 0.9275

This entry is a complete valid challenge entry.