removing rows contain NaN element from 3D array

A(:,:,1) =

                 NaN                 NaN                 NaN   1.000000000000000
                 NaN                 NaN                 NaN   1.000000000000000
                 NaN                 NaN                 NaN   1.000000000000000
  -0.251395732283393   0.831232695305901   0.495835045195662   1.000000000000000
  -0.351395732283393   0.831232695305901   0.495835045195662   1.000000000000000
                 NaN                 NaN                 NaN   1.000000000000000
   0.223016679421216   0.961124117481440  -0.162800465280223   1.000000000000000

A(:,:,2) =

                 NaN                 NaN                 NaN   1.000000000000000
                 NaN                 NaN                 NaN   1.000000000000000
                 NaN                 NaN                 NaN   1.000000000000000
  -0.294547842132722   0.762119326903777   0.576555027950231   1.000000000000000
                 NaN                 NaN                 NaN   1.000000000000000
   0.208731731285917   0.936100623203798  -0.283101903193612   1.000000000000000
                 NaN                 NaN                 NaN   1.000000000000000

How can I remove rows contain NaN from 3D A array? After removing each row contain NaN, sub-arrays' dimensions will not be equal. Is it allowed in Matlab?

Answer 1

[Guillaume]

No, you cannot have different size pages in a matrix.

You could convert the 3D array into a cell array of 2D matrices and remove the nans from these matrices:

cellA = num2cell(A, [1 2]); %keep dimension 1 and 2 together
nonancellA = cellfun(@(m) m(~any(isnan(m), 2), :), cellA, 'UniformOutput', false);

Whether or not it makes later processing easier, only you can say.

Answer 2

[Andrei Bobrov]

out = arrayfun(@(x)A(all(~isnan(A(:,:,x)),2),:,x),1:size(A,3),'un',0);

Answer 3

[KSSV]

You can take the output in cell.

[m,n,p] = size(A) ;
iwant = cell(p,1) ;             
for i = 1:p
    % get Nan's from first column 
    temp = A(:,1,i) ;
    id = ~isnan(temp) ;
    iwant{i} =A(id,:,i) ;       
end

You can access the required matrix using iwant{1}, iwant{2}.