Matlab MLTOOLS Toolbox

The MLTOOLS software can be downloaded here.

Release Information

Current release is 0.126.

As well as downloading the MLTOOLS software you need to obtain the toolboxes specified below.

Toolbox	Version
OPTIMI	0.132
NDLUTIL	0.156

Modified kbrParamInit to scale alpha weights and biases by number of data. Added 'dynamicsSliderChange' to lvmClassVisualise to allow visulisation of models with 'gpTime' style-dynamics.

Version 0.125

Added multimodel for learning multiple indepedent models with shared parameters.

Version 0.124

Added periodic RBF network and model gradient checker.

Version 0.123

Minor release in line with IVM toolbox 0.4.

Version 0.122

Added Hessian code for base model type and for MLP. Added Viterbi alignment code, viterbiAlign.

Version 0.121

Various minor bug fixes and changes which seem to have gone undocumented.

Version 0.12

Extended model type to be a generic container module for optimising any model. Added model test for testing a created model. The code is still in a bit of flux though with some design decisions not made and some code untested.

Version 0.111

Fixed bug in kbr where bias parameter fields where still being referred to as b.Also acknowledged the fact that the isomap graph may not be fully connected in isomapEmbed, but don't yet deal with it properly. Finally added lleEmbed.m for wrapping the lle code.

Version 0.11

Updated release for operation with FGPLVM toolbox 0.13. Structure of model creation changed and functions of the form modelOptions.m included for setting default options of the various machine learning models.

Version 0.1

The first release of the toolbox with various wrappers for NETLAB functions. Also latent variable model visualisation code was moved into this toolbox.