## SOM output labelling

3 posts
• Page

**1**of**1**### SOM output labelling

Dear Orange Community,

I am using the Self-Organising Maps function which is so much better than simple clustering as you get to know more about the "why?" aspect. However, I am a little unsure about the values in the layer labelling.

I am choosing one particular layer (component plane) in the SOM Visualizer and I get a nice grey scale from 0.03 - 21.84. When I look at the extremes of the data for that component, in the raw data it ranges between 0.00 and 42.13.

So, what is the scale a measure of?

Thanks

Steve

I am using the Self-Organising Maps function which is so much better than simple clustering as you get to know more about the "why?" aspect. However, I am a little unsure about the values in the layer labelling.

I am choosing one particular layer (component plane) in the SOM Visualizer and I get a nice grey scale from 0.03 - 21.84. When I look at the extremes of the data for that component, in the raw data it ranges between 0.00 and 42.13.

So, what is the scale a measure of?

Thanks

Steve

### Re: SOM output labelling

smmudge wrote:I am choosing one particular layer (component plane) in the SOM Visualizer and I get a nice grey scale from 0.03 - 21.84. When I look at the extremes of the data for that component, in the raw data it ranges between 0.00 and 42.13.

So, what is the scale a measure of?

SOM Visualizer shows the scale for the 'codebook' vectors (coordinates of SOM nodes in the data space). They generally lie inside the data, so their ranges are smaller, and depend on the data density distribution.

### Re: SOM output labelling

Ales, thanks for the reply. Can I make an analogy between the codebook vectors and the upper and lower quartile limits in a box and whisker plot by any chance with the unusual values being outliers and therefore with a low data density distribution? Steve

3 posts
• Page

**1**of**1**