0
$\begingroup$

I wish to normalize inputs parameters into [0 - 1] and fit into a neural network for training. I have done a simple normalization method - MinMax

Question 1: if my inputs have negative values, do i need to increase each of the values by adding with the min negative value's absolute value? So there will not be any negative value left.

Example training sets.

enter image description here

Question 2: Do i have to normalize my outputs with my inputs as well? (together)

$\endgroup$

1 Answer 1

1
$\begingroup$
  1. Yes.
  2. No, but it might help learning.

When you use minmax normalisation, make sure that your input domain is bounded, i.e. that now values outside of that interval can occur. Otherwise, subtract the mean and divide by the standard deviation.

$\endgroup$
2
  • $\begingroup$ Thanks for your reply, i believed i have to normalize each row - refer to the image. Such that the min for row 1 is -0.1 and min for row 2 is 1.0. Am i right? $\endgroup$ Commented Oct 20, 2014 at 10:22
  • $\begingroup$ No, you normalise columns in your case. $\endgroup$
    – bayerj
    Commented Oct 20, 2014 at 11:27

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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