# Remove rows or cols whose elements are all NaN

877 views (last 30 days)
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.

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),:)=[];
Walter Roberson on 20 Sep 2015
Michal Gajewski commented
works

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

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

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

Jeffery Devereux on 20 Jun 2017
Thank You!
yonca on 7 Oct 2019
thank you so much

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

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

Jan on 25 Mar 2013
This does not solve the wanted: "delete rows or cols whose elements are all NaN"
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 on 4 Apr 2016
Your answer will create an array. he wants to keep the matrix.

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"

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.

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

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 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.