I'm training the self organizig map, I need to set the value for the radius of it. is there any method to find the optimum radius size ?


It depends on the range and scale of your input data. If you are mean-zero standardizing your feature values, then try $\sigma=4$. If you are normalizing feature values to a range of [0,1] then you can still try $\sigma=4$, but a value of $\sigma=1$ might be better. Remember, you have to decrease the learning rate $\alpha$ and the size of the neighborhood function with increasing iterations, as none of the metrics stay constant throughout the iterations in SOM.

It also depends on how large your SOM is. If it's a 10 by 10, then use for example $\sigma=5$. Otherwise, if its a 100 by 100 map, use $\sigma=50$.

In unsupervised classification, $\sigma$ is sometimes based on the Euclidean distance between the centroids of the first and and second closest clusters.

  • $\begingroup$ Interesting, I've been using SOM for different applications and always stuck with the default parameters. The values you mention look like rules of thumb. Could you please provide some reference(s) to learn more? $\endgroup$ – Vladislavs Dovgalecs Mar 31 '15 at 15:48
  • $\begingroup$ Look at Kohonen's book on SOMs. $\endgroup$ – JoleT Apr 3 '15 at 14:40

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.