I want to understand the implementation of SOM using MATLAB. I tried the iris flowers clustering demo program. I am not able to understand the hit diagram. Also once the map is trained how to test it?
please go through basic paper by kohonen for understanding. For SOM testing you simply simulate it with new data set and see the classes it throws, match it for exact classification. Make TP, TN, FN,FP and check its accuracy.
Dear Rajiv,
I had tried the iris flower example available in MATLAB demo but am not able to understand the plots I am getting. It would be really useful if you can throw some more light.
Dear Aarti,
I recommend the SOMToolbox from the Aalto-University Helsinki. This Toolbox is much better than the standard implementation inj MATLAB and provides many additional tools for analysis of the trained SOM. There are also several demos provided together with this toolbox. To get them all, simple google for 'SOMToolbox Helsinki'
Best, Thomas Villmann
I actually have got similar problem. I have trained my network and generated simulink, but when I'm trying to test it, it comes up with an error.
I have tried example 'iris' using SOM but the dataset is RGB and i have also trained my data set over SOM, but i'm still confuse with structure like we give dimension 1 and dimension 2, why we give these two dimension.
Second thing is if we want some to get suitable output from SOM which we can further use therefore what will be suitable output 'a image', 'feature vector'?
- Badges
- Science topic