Visualisation Induced SOM (ViSOM)
When used for visualisation of high dimensional data, the self-organising map (SOM) requires a colouring scheme such as U-matrix to mark the distances between neurons. Even so, the structures of the data clusters may not be apparent and their shapes are often distorted. In this paper, a visualisation-induced SOM (ViSOM) is proposed as a new tool for data visualisation. The algorithm constrains the lateral contraction forces between a winning neuron and its neighbouring ones and hence regularises the inter-neuron distances. The mapping preserves directly the interneuron distances on the map along with the topology. It produces a graded mesh in the data space and can accommodate both training data and new arrivals. The ViSOM represents a class of discrete principal curves and surfaces.
KeywordsData Space High Dimensional Data Winning Neuron Data Structure Analysis Nonlinear Principal Component Analysis
Unable to display preview. Download preview PDF.
- .Kohonen, T., Self-Organising Maps, Springer: Berlin, 1995.Google Scholar
- .Ultsch, A., “Self-organising neural networks for visualisation and classification,” in Information and Classification, O. Opitz, B. Lausen and R. Klar eds., pp. 864-867, 1993Google Scholar
- .Fisher, R.A., “The use of multiple measurements in taxonomic problems,” Annual Eugenics, vol. 7, pp. 178-188, 1936.Google Scholar