calculate distance matrix with Tanimoto function

KNIME Analytics Platform
#1

Hello,

Is this possible to generate a distance matrix using Tanimoto function for simple double data?
I have a such matrix, normalized between 0 and 1, and try to do that. It works with Euclidean and Manhattan functions, but when I select Tanimoto,
my columns disappear from the windows selection...
I think it is a problem of data type (I read on the forum that lot of users use this node to calculate similarity with fingerprints), but could you confirm?

In that case, is this reasonable to image a Tanimoto coefficient that gives information of similarity between doubles?

Nicolas

#2

The Tanimoto coefficient is defined only for bitvectors (and fingerprints are just  bitvectors).

#3

I think, the question is about the visualization of the similarities.
I have the same problem here.

I calculated pairwise tanimoto coefficients and now I have a vector for each pair of bitvectors. Now, I would like to visualize it.

 

The question is how can you process a column with vectors most efficiently and also visualize it?

 

best Soren

#4

I think the Heatmap (JFreeChart) node is what you are looking for. (See page 26, 27 in the book sample/preview for example image.)