Normalizing data

Hi there,

How would you normalize a column of data in matlab between 0 and 1 using the formula, z = (x-mean)/standard deviation. Is there a built in function to do this?

Many thanks

John

Answer 1

Dear John,

Lets say you have matrix D and you want to normalize each value of Column to unit length (between 0-1). One possible way is :

D = bsxfun(@rdivide,D,sum(D));

each column will be unit normalized. If you take the sum of each column it will be one.

sum(D)

Do let me know if I understand you correct.

zscore from the statistics toolbox does (x-mean)/standard deviation

Answer 2

Dividing by the maximum does not correct for offset. So only ranges 0-X go to 0-1. If his column has a negative or nonzero minimum it won't normalize to 0-1.

While the formula for correcting this is simple, mat2gray will normalize to 0-1 for you:

for i = 1:ncol
D(:,i) = mat2gray(D(:,i))
end

You have to decide if you need to normalize before or after your calculation.