5
$\begingroup$

I generate input data for my model at different temperatures. At each temperature I generate 1000 samples and in each sample I have 16 features. So the shape of array is (1000,16) at each temperature. I want to normalize data between 0 and 1. But I am really confused that what is the strategy of normalization in machine learning language? Should I normalize feature of each single sample such that each sample is like [0,.....1] or should I normalize along column axis. Secondly I also want to know that should I perform this normalization separately for each temperature?

$\endgroup$
6
  • $\begingroup$ You normalize each variable (column) separately. $\endgroup$ Commented Oct 25, 2018 at 8:00
  • $\begingroup$ You mean feature of every single sample? $\endgroup$
    – herry
    Commented Oct 25, 2018 at 9:22
  • $\begingroup$ You normalize each variable (feature) separately, for all samples. $\endgroup$ Commented Oct 25, 2018 at 9:26
  • $\begingroup$ Sorry may be I'm misunderstanding your statement.In my question I have 16 features in each sample. So after normalization I will get [0......1].length of this vector is 16.Is this what you want to convey? $\endgroup$
    – herry
    Commented Oct 25, 2018 at 10:28
  • $\begingroup$ Normalize the whole feature (with all the samples), that is extract every 1000 length vector (total of 16), apply normalization to each such vector. $\endgroup$ Commented Oct 25, 2018 at 10:32

1 Answer 1

2
$\begingroup$

You should normalize each column separately. After normalization each feature should have a maximum value at 1 and a minimum at 0.

$\endgroup$

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.