Re: [Databionic-ESOM-User] U*-Matrix

[Fabian M. <fa...@my...>]

Amphun,

as far as I know, the U*Matrix has not been implemented in
the Java tools but you can get a Matlab version from Prof Ultsch.

What do you mean with validation? Since it's an unsupervised
method you could use the quantization error to measure the
approximation quality. You also use a method to obtain clusters
from the map and use cluster validity mearues. You could use
a class mask to classify unseen data and measure classification
accuracy. Finally you could compare maps to measure the
reprocability. What do you want to measure?

best
fabian

Amphun Chaiboonchoe wrote:
> Dear All,
>
> I am new for ESOM, I start analyze my data with ESOM.
> I wonder how can I see the U*-Matrix, at visualization tool I only found U-Matrix and P-Matrix.
>
> Another question is, which is the best validation index for ESOM. Does anyone try to validate 
> ESOM before ? Can you share more info please ?
>
> Thank you in advance.
> Regards,
> Amphun.
> -----------------------------------------------------
> Amphun Chaiboonchoe
> PhD Student 
> Room No.7 Hurunui Building
> Environment, Society and Design Division
> PO Box 84, Lincoln University
> Lincoln 7647, Canterbury
> New Zealand
> Phone: +64-3-3253838 Extn. 8811
> Mobile: 021-02454425
>
>
>
>
>