Welcome to the GTM homepage

GTM, which stands for generative topographic mapping, is a model for density modeling and data visualisation, developed at the Neural Computing Research Group. This page is meant to provide information about the results from this research, in terms of papers, technical reports, software, demos etc.

Publications available on the GTM

The currently most complete desription of the GTM can be found in

GTM: The Generative Topographic Mapping, PhD thesis by Markus Svensén, 1998. Available as compressed postscript, abstract available on-line; approx. 110 pages, 44 figures. (BibTeX entry)

In addition, the following papers are available in compressed postscript format. (Abstracts available on-line.)

Developments of the GTM, to appear in Neurocomputing, special issue on the Self-Organizing Map. (BibTeX entry)
GTM: The Generative Topographic Mapping, Neural Computation 10(1). (BibTeX entry)
Magnification factors for the SOM and GTM algorithms, presented at WSOM'97. (BibTeX entry)
GTM: A Principled Alternative to the Self-Organizing Map, presented at NIPS*96. (BibTeX entry)

Software for GTM

A MATLAB(R) implementation is available. Alternative implementations are being worked on, but no releases are planned for the near future.

Data Sets

As we try different data sets in experiments with the GTM, we will try to make these data sets available via this page; at the moment, there are only two.

Measurements from a multi-phase flow pipe line.

Leptograpsus crabs on Prof. Brian Ripley's pages at Oxford.

Demos and Graphics

This will be a place for various visual bits that illustrate different aspects of the GTM. It's still early days so all there is at the moment, is a little mpeg movie that shows the GTM learning a toy problem in 2D.

Toy data demo.

Pages: NCRG - Who's who - Research - Publications - Resources - Aston

This page is maintained by Markus Svensén (svensen@cns.mpg.de)