I have a theoretical question: does the Normalizer node take into consideration the negative values for discrete variables? In my data I have discrete numbers and the sign has a peculiar meaning (similar to the direction in a vector). I could not find a description of the background math for this node, but the general formula for the min-max normalization should take the sign in account. Can someone confirm if the node does it as well?

Hi Silvia
the normalization will take the sign into account. Hence if you map a interval from -10 to 1 to a 0 to 1 normalization afterwards -10 will be represented as 0. All other values are distributed inbetween.

Thank you Iris, I wanted to be sure Do you think it is important for a clustering algorithm to differentiate those values, if they have a different meaning? What I mean is: if I want to have entries with negative descriptors in different clusters than entries with positive descriptors, should I normalize them separately or the min-max will do the job?
Thanks again and hope to hear from you soon!

From your description, I think you should consider using z-score output in the normalizer instead. That is going ti provide you the measure of the distance from the mean in STDDEV units where negative numbers are to the left of the mean and positive numbers become to the right of the mean. I usually use z-scores in models for my clients.