Select a random cell from multiple matrices

let's say 3 matrices,

A = 30x20; B = 30x20; C = 30x20;

All three matrices contain NaNs.

I want to select a random value from each row out of ALL three matrices but none of the corresponding values should be a NaN. For e.g., if the random value is A(1,10) then B(1,10) and C(1,10) should not contain a NaN. Also it has to be the same cell number(row,column) for all matrices. If nothing matches moves to the next row.

And then repeat it for each row until the end.

[Purushottama Rao]

What is there inside A B C? does they contain Nan in few positions? What do you expect as output? is it a 30*20 matrix containing 1 s and 0 s?

[Geant Bepi]

A, B and C matrices contain just random numbers (integers). Nans are just all over.

I expect the random value as the output.

Answer 1

[Guillaume]

If you want to go randomly through all the elements of the matrices for which none of the matrices contain a nan, simply use randperm to reorder the linear indices that do not contain NaN in any of your matrices:

validindices = find(~(isnan(A) | isnan(B) | isnan(C))); %linear indices that have no nans in any of the matrices
randindices = validindices(randperm(numel(validindices))) %reorder randomly

If you want subscripts instead of linear indices, just use ind2sub to convert, although you can use the linear indices just fine:

[randrows, randcols] = ind2sub(size(A), randindices);