This page describes examples of how to use ensemble learning in neural networks software (ENSMLP). The software includes variational approximations that are Gaussian and mixtures of Gaussians used to approximate neural network posterior distributions.
Current release is 0.1.
As well as downloading the ENSMLP software you need to obtain the toolboxes specified below. These can be downloaded using the same password you get from registering for the ENSMLP software.
First release in response to a request for the code. Code was written in 1998-1999 but is being released for first time in 2007. The code is heavily based on the NETLAB toolbox, to such an extent that copyright from NETLAB probably applies to large portions of this software. Please see GPL licenses on that software for details of the implications of this.
The ensemble learning is demonstrated with a series of examples on the 'Tecator' data of Thodberg.
There are several different configurations of the models to run on the
Tecator data, all start with '
dem'. I've put together this code
release over a couple of days, and haven't managed to recreate exactly
the results on this data we quote in our original tech report. Be aware
also that the scripts each take a few hours to run. Finally the
demTecatorMixEns... scripts can only be run once the corresponding
demTecatorEns... script has been run.
Finally there is a Gaussian process demo,
demTecatorGpRbfArd that you
will need to download my GP toolbox to run.