0
$\begingroup$

I have binary data input to NN. The weights generated by NN are not normalized, i.e., the weights are not the same for every run of the algorithm. How to standardize these weights?

$\endgroup$
0
$\begingroup$

This is not a normalisation problem, it's an artifact of random initialisation of the weights. A simple solution is to run training a few times and pick the weights that minimise your loss function over the different runs.

$\endgroup$
1
  • $\begingroup$ Thank you. But can we automate this procedure of finding the best run $\endgroup$ – user857418 Mar 3 '14 at 6:15

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.