1
$\begingroup$

First of all, is it good to do normalization and standardization for Radial basis neural network?

Does MATLAB do that automatically for RBF? I've read that MATLAB does normalization and standardization for Feedforward neural network, but I am not sure If it does for RBF too. I use nntool and I don't see if MATLAB does normalization or not.

$\endgroup$
1
$\begingroup$

Range normalization (0,1) is necessary for RBF network. MATLAB doesn't normalize RBF networks automatically (only feedforward network).

| cite | improve this answer | |
$\endgroup$
  • $\begingroup$ Radial basis function networks are normalized on the interval {0,1} since they use positive norms to do so, n'est pas? Answer could use a bit more text to be more robust. $\endgroup$ – Carl Oct 14 '16 at 1:23
  • $\begingroup$ Is it actually mistake use {-1,1} normalization instead of 0,1? Gaussian function in RBF should makes numbers correct anyway. $\endgroup$ – EdWood Oct 14 '16 at 2:49
  • $\begingroup$ If you think the normalization would be {-1,1}, then find a reference to that effect. I gave a reference for the {0,1} interval because that is what I found. I have no idea what it is used for. Maybe you should make that clear in your answer. $\endgroup$ – Carl Oct 14 '16 at 5:27
  • $\begingroup$ I think it doesn't matter which size of range you set until you set it for all atributes. I do calculation for it later. You posted link on wiki, but I dont see any mention about normalization on the interval {0,1}. $\endgroup$ – EdWood Oct 21 '16 at 15:39
  • $\begingroup$ It is not stated explicitly. However, as $\rho>0$ and the normalization uses $\rho$ times the ratio of absolute values or their sums, then the interval is {0,1}. $\endgroup$ – Carl Oct 21 '16 at 17:29

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.