# Interpreting a self organizing map resulting from the same dataset with different standard deviations

I am working on a task where the goal is to make a sofm learn a mapping from a three-dimensional space (the input space) to a two-dimensional space (the sofm "grid").

The data points are drawn from two different probability distributions, F1 and F2, forming two different clusters. There are two data sets, P10, and P30, each with 100 points drawn from F1 and 100 points drawn from F2. The two clusters are centered around the same points in both data sets, but they have a different amount of spread. The name of each data set indicates the standard deviation of the angle between each data point in the set and the center of the cluster that it is from. The standard deviations are 10°, and 30°, respectively.

first I trained a SOM using P10 data-->SOM_P10 and another SOM using P30 data-->SOM_P30. Both SOMs have the same configuration: som = newsom(P10, [10 10], 'hextop', 'linkdist', 100, 5) and som = newsom(P30, [10 10], 'hextop', 'linkdist', 100, 5) and this is the resulting plotsomhits (which shows which nodes are the winners) for cluster F1:  Later I tried to play around with the data. So I trained SOM_P30 with P10 and SOM_P10 with P30. And these are the resultant plotsomhits for cluster F1:  As you can see nothing has changed. And I don't know how can I explain this !

• Could you please put your main specific question in the title, just to clarify what it is?
– nbro
Mar 3 at 9:07