Remove rows or cols whose elements are all NaN

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]

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]

Michal Gajewski commented

works

Answer 1

[Andrei Bobrov]

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

[Jan]

A can be a row-vector, such that all will operate along the 2nd dimension automatically. Therefore this is safer:

out = A(:, all(~isnan(A), 1));

[HUGO LOYA]

Why do you write 1 for columns and 2 for rows?

[Jan]

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

[Céldor]

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]

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"

Answer 2

[Phillippe]

To remove only ALL-NaN columns, do this instead:

A = A(:,~all(isnan(A)));

[Jeffery Devereux]

Thank You!

[eda]

thank you so much

[Parth Dev Bundela]

Thanks man

[osman alper altun]

Thank you!

Answer 3

[Saman]

Use this :

   out = A(:,any(~isnan(A)));  % for columns
   out = A(any(~isnan(A),2),:);   %for rows

Answer 4

[Azzi Abdelmalek]

A(isnan(A))=[]

[Jan]

This does not solve the wanted: "delete rows or cols whose elements are all NaN"

[Azzi Abdelmalek]

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]

Your answer will create an array. he wants to keep the matrix.

[Serafeim Zacharopoulos]

To delete rows with all NaN's, maybe try this:

A = A(find(sum(~isnan(A)'))',:)

[Walter Roberson]

A(all(isnan(A),2),:) = [];    %rows that are all nan
A(:, all(isnan(A),1)) = [];   %cols that are all nan

[Andrew Sol]

You proposed too simple a solution and you were wrong.

Answer 5

[Alfaz Memon]

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]

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]

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.

Answer 6

[Ilham Hardy]

Haven't tried this, but it should works:

A(isnan(A))=[];

[Jan]

This does not solve the wanted: "delete rows or cols whose elements are all NaN"

Answer 7

[Nike]

Simplest is

A(isnan(A))= [];

[Jan]

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.