0
$\begingroup$

I want to implement a neural network in Pytorch for medical image segmentation. I should normalise my data.

Should I apply a min-max scale (range 0 to 1) before applying the normalisation or should I apply the z-score normalisation directly?

  1. What is the best approach if the dataset comes from a single source of data?

  2. What if the samples come from multiple sources?

Improve this question
$\endgroup$

1 Answer 1

Reset to default
2
$\begingroup$

There is, beside of numerical losses not difference between directly using z-normalization and first min-max and then z-normalization.

Explanation

Both are affine transformations and a combination of affine transformations is again an affine transformation. Since there is only one(*) affine transformation that results on zero-mean and variance of one, the z-normalization will (kind of) undo all previous affine transformations to achieve this goal.

(*): In fact there are two and you could get the other by multiplying every value with -1. But this is just a silly detail and does not matter for this question.

Improve this answer
$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .