adding distance functions to existing algorithms

Hi there,

is there a way to add new distance functions to existing algorithms (like k-means, hirachical clustering, knn, etc).

I am dealing with multivariate time series data, where an euclidian distance does not make much sense. We have some ideas that use dynamic time warping or the shapes of the curve or compression to compare two entities. We are considering to implement those in knime to use them for clustering/classification. However I dont see the point to re-implement the say k-means algorithm (which i would need if i create a new node, wouldn't I?).

Do you have any suggestions?

regards, d


Have you seen this presentation yet? According to it, the org.knime.distmatrix extension point might be the one you need.

Cheers, gabor

In org.knime.distmatrix there is an extension point org.knime.distmatrix.DistanceFunctions. This allows you to add additional distance functions from other plug-ins. This function can then be used in the Distance Matrix Calculator and the resulting distance matrix in any node that works in distance matrices.