Assessing self organizing maps via contiguity analysis

- Contiguity Analysis is a straightforward generalization of Linear Discriminant Analysis in which the partition of elements is replaced by a more general graph structure. Applied to the graph induced by a Self Organising Map (SOM), Contiguity Analysis provides a set of linear projectors leading to a representation as close as possible to the SOM map. As expected, such projectors may only concern local parts of the SOM maps. They allow us to visualize the shapes of the clusters and the pattern of the elements within each clusters. In some contexts, they provide confidence areas for elements via a standard partial bootstrap procedure. Key words

• Type: Journal
• Year: 2006
• Where: NN
• Authors: Ludovic Lebart
• Keywords: Contiguity Analysis, General Graph Structure, Linear Discriminant Analysis, Neural Networks, NN 2006