MATLAB Answers

2

How to normalize values in a matrix to be between 0 and 1?

Asked by Sahar abdalah on 8 Apr 2015
Latest activity Commented on by Tubi
on 22 Mar 2018

I have a matrix Ypred that contain negative values and I want to normalize this matrix between 0 and 1

 Ypred=[-0.9630 -1.0107 -1.0774
       -1.2075 -1.4164 -1.2135
       -1.0237 -1.0082 -1.0714
       -1.0191 -1.3686 -1.2105];

I'm new in matlab, please help me, there is a matlab function or toolbox that can do this? thanks

  2 Comments

What do you mean by "normalize"? Divide by the max value in the matrix and make all values positive? Do this by columns or rows? Divide by norm of columns or rows? Or what?

I want to divide by norm of row to make positive values

Sign in to comment.

4 Answers

Answer by Jos (10584)
on 8 Apr 2015
Edited by Jos (10584)
on 8 Apr 2015
 Accepted Answer

This can be simply done in a two step process

  1. subtract the minimum
  2. divide by the new maximum
normA = A - min(A(:))
normA = normA ./ max(normA(:)) % *

note that A(:) makes A into a long list of values. Otherwise min(A) would not return a single value ... Try fro yourself!

  • Edited after comment ...

  2 Comments

Thanks for the code, I just tried this and it not normalised between 0 and 1

Ypred =

   -0.9630   -1.0107   -1.0774
   -1.2075   -1.4164   -1.2135
   -1.0237   -1.0082   -1.0714
   -1.0191   -1.3686   -1.2105

normA =

    0.4534    0.4057    0.3390
    0.2089         0    0.2029
    0.3927    0.4082    0.3450
    0.3973    0.0478    0.2059

normA =

    1.0000    1.0495    1.1188
    1.2539    1.4708    1.2601
    1.0630    1.0469    1.1126
    1.0583    1.4212    1.2570

Sorry! The second line of code is wrong ;-) It should read

normA = normA ./ max(normA(:))

Sign in to comment.


Answer by James Tursa
on 8 Apr 2015

NormRows = sqrt(sum(Ypred.*Ypred,2));
Ynorm = bsxfun(@rdivide,abs(Ypred),NormRows);

  3 Comments

If I want to use this code for my work, any way should I use to cite/reference?

Why would you need a citation? If you truly want a citation, just cite MATLAB itself, or perhaps the doc page for bsxfun, but it seems a bit silly to need a citation for a simple code fragment.

Many Thanks John, I believe you are right in your suggestion since they are mostly common MATLAB commands and functions.

Sign in to comment.


Answer by Sahar abdalah on 9 Apr 2015

thank you for your answers. I used both codes and I found two different result. what is the result that I can use?

    normA = Ypred - min(Ypred(:))
    normA = normA ./ max(normA(:))
    normA =
    1.0000    0.8948    0.7477
    0.4607         0    0.4475
    0.8661    0.9003    0.7609
    0.8763    0.1054    0.4541
   NormRows = sqrt(sum(Ypred.*Ypred,2));
   Ynorm = bsxfun(@rdivide,abs(Ypred),NormRows);
   Ynorm =
    0.5461    0.5731    0.6110
    0.5435    0.6375    0.5462
    0.5712    0.5625    0.5978
    0.4871    0.6542    0.5786

  2 Comments

Use whichever is appropriate for your problem. Jos and I are both trying to interpret what you want, but without any background about the problem you are solving it is a bit of a guessing game at our end.

Sign in to comment.


Answer by Candan Celik on 12 Aug 2017

Hello, What is the name of this norm in the accepted answer? (Euclidean, Frobenius etc.)

  1 Comment

Please post comments to answers in the section for the comments. You message is not an answer.

The accepted answer does not contain a norm at all, but a "normalization". A matrix norm would reply a scalar, the normalization replies a matrix with the same size, but with shifted and scaled values.

Sign in to comment.


Discover what MATLAB® can do for your career.

Opportunities for recent engineering grads.

Apply Today