MATLAB Answers

Jeon
3

Remove rows or cols whose elements are all NaN

Asked by Jeon
on 25 Mar 2013
Latest activity Commented on by 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

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),:)=[];
Michal Gajewski commented
works

Sign in to comment.

7 Answers

Answer by Andrei Bobrov
on 25 Mar 2013
Edited by Andrei Bobrov
on 25 Mar 2013
 Accepted Answer

out = A(all(~isnan(A),2),:); % for nan - rows
out = A(:,all(~isnan(A))); % for nan - columns

  5 Comments

Jan
on 13 Jul 2014
@Hugo: See doc all for details. The numbers specify the dimension to operate on.
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
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.


Answer by Phillippe on 14 Jan 2015

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

  2 Comments

thank you so much

Sign in to comment.


Answer by Saman
on 23 Oct 2013
Edited by 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.


Answer by Azzi Abdelmalek
on 25 Mar 2013

A(isnan(A))=[]

  3 Comments

Jan
on 25 Mar 2013
This does not solve the wanted: "delete rows or cols whose elements are all NaN"
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.
Your answer will create an array. he wants to keep the matrix.

Sign in to comment.


Answer by Ilham Hardy on 25 Mar 2013

Haven't tried this, but it should works:
A(isnan(A))=[];

  1 Comment

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

Sign in to comment.


Answer by Nike
on 25 Mar 2013

Simplest is
A(isnan(A))= [];

  1 Comment

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.


Answer by Alfaz Memon on 20 Aug 2018
Edited by 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

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.
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.