MATLAB Answers

Remove rows or cols whose elements are all NaN

877 views (last 30 days)
Jeon
Jeon on 25 Mar 2013
Commented: yonca on 7 Oct 2019
How can I remove rows or cols whose elements are all NaN ? Withouot any dirty iterations?
For example,
A = [1 1 1 1 1 1 1 1 1 1;
NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN;
1 1 1 1 1 1 1 1 1 1;
NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN;];
should turned into
A = [1 1 1 1 1 1 1 1 1 1;
1 1 1 1 1 1 1 1 1 1];
This is just an example. Actually I have a very big matrix. So I want a solution of this question to work well with my big matrix.

  2 Comments

Dev-iL
Dev-iL on 13 Jul 2014
Hi! You can find some clues here: A discussion of the opposite problem on SE. I personally achieved what you were attempting using:
A(~any(~isnan(A), 2),:)=[];

Sign in to comment.

Accepted Answer

Andrei Bobrov
Andrei Bobrov on 25 Mar 2013
Edited: Andrei Bobrov on 25 Mar 2013
out = A(all(~isnan(A),2),:); % for nan - rows
out = A(:,all(~isnan(A))); % for nan - columns

  5 Comments

Show 2 older comments
Jan
Jan on 13 Jul 2014
@Hugo: See doc all for details. The numbers specify the dimension to operate on.
Céldor
Céldor on 2 Apr 2015
Can I obtain a logical matrix I of indices to be held / removed something like
I = something here
and then use assign an altered matrices:
out1 = out1(I);
out2 = out2(I);
% ... etc.
Thanks
gringer45
gringer45 on 18 Mar 2018
This doesn't really do what the question asks for. This selects all the columns or rows with none (zero) NaN values. So, this is answering the question: "Remove rows or cols whose elements have any (at least one) NaN"

Sign in to comment.

More Answers (6)

Phillippe
Phillippe on 14 Jan 2015
To remove only ALL-NaN columns, do this instead:
A = A(:,~all(isnan(A)));

Saman
Saman on 23 Oct 2013
Edited: Saman on 23 Oct 2013
Use this :
out = A(:,any(~isnan(A))); % for columns
out = A(any(~isnan(A),2),:); %for rows

  0 Comments

Sign in to comment.


Azzi Abdelmalek
Azzi Abdelmalek on 25 Mar 2013
A(isnan(A))=[]

  3 Comments

Jan
Jan on 25 Mar 2013
This does not solve the wanted: "delete rows or cols whose elements are all NaN"
Azzi Abdelmalek
Azzi Abdelmalek on 25 Mar 2013
I am not sure what he want, from his example his rows are all 1 or all nan! In this case the answer solve his problem.
Emanuel-Petre Eni
Emanuel-Petre Eni on 4 Apr 2016
Your answer will create an array. he wants to keep the matrix.

Sign in to comment.


Ilham Hardy
Ilham Hardy on 25 Mar 2013
Haven't tried this, but it should works:
A(isnan(A))=[];

  1 Comment

Jan
Jan on 25 Mar 2013
This does not solve the wanted: "delete rows or cols whose elements are all NaN"

Sign in to comment.


Nike
Nike on 25 Mar 2013
Simplest is
A(isnan(A))= [];

  1 Comment

Jan
Jan on 25 Mar 2013
This has been posted twice already. But it still does not solve the original question:
delete rows or cols whose elements are all NaN
For e.g. A = [1, NaN, 1; NaN, 1, NaN] nothing should be deleted.

Sign in to comment.


Alfaz Memon
Alfaz Memon on 20 Aug 2018
Edited: Alfaz Memon on 21 Aug 2018
input varibale : data
output variable : row_index( index of rows with all the value as NaN)
row_index( index of rows with all the value as NaN)
column_index( index of columns with all the value as NaN)
if(iscell(data))
x =find(cell2mat((cellfun(@(data) any(isnan(data),2),data,'UniformOutput',false))));
[ia,ib] = ind2sub(size(data),x);
rows_unique = unique(ia);
rows_unique(:,2)=histc(ia,rows_unique);
row_index = rows_unique(find(rows_unique(:,2)==size(data,2)),1);
columns_unique = unique(ib);
columns_unique(:,2)=histc(ib,columns_unique);
column_index = rows_unique(find(columns_unique(:,2)==size(data,1)),1);
else
row_index =find(~any(~isnan(data), 2)); % row with all NaN values
column_index =find(~any(~isnan(data), 1)); %column with all NaN values
end

  2 Comments

Walter Roberson
Walter Roberson on 20 Aug 2018
Seems like a bit of a bother to just remove the rows or columns ?
I notice that you are using cellfun on the data, implying that the data is a cell array; in the original question it was a plain array.
Alfaz Memon
Alfaz Memon on 21 Aug 2018
yeah its for cell array. for plain array you can remove cellfun and just simply keep any(isnan(data),2) instead of x =find(cell2mat((cellfun(@(data) any(isnan(data),2),data,'UniformOutput',false))));
this can you work for cell array of different data type.
Also I have updated solution.

Sign in to comment.

Sign in to answer this question.