How to assign a NAN to an empty cell in a cell array of cell array of matrix

I want to set NAN to these cells. Basically i wanted to do a basic threshold operation on a cell array of cell array of matrix . Some of the cells are empty cells, while others have values. If the absolute values in the cells are less than a threshold value say 5, i want to assign a zero to it. The problem is that the empty cells are also set to zero, which i do not want. So if we convert empty cell to NAN this problem can be solved. After the operation i can reassign NAN as an empty cell. But I am not able to perform this operation. I am able to do it in the first level based on previous discussion in the forum. for eg:

c={[] [] [2 4 5] [] [2 3 4 6 7]}%  1×5 cell array -- []    []    [1×3 double]    []    [1×5 double]
fx=@(x)any(isempty(x))
ind=cellfun(fx,c)
c(ind)={nan}

I get

c =
  1×5 cell array
    [NaN]    [NaN]    [1×3 double]    [NaN]    [1×5 double]

Now for this example how can i do the same process?

c={{[] [1 2]} { [] [] [2 4 5]} {[] [2 3 4 6 7] []}}

c =
  1×3 cell array
    {1×2 cell}    {1×3 cell}    {1×3 cell}

When i follow the previous step I am getting the error "Function 'subsindex' is not defined for values of class 'cell'."

[James Tursa]

Please also show the code you are currently using to do the threshold testing & 0 assigning.

[GEEVARGHESE TITUS]

I have just written a coarse threshold function and applied to my data using cellfun.

    function y = sthresh(x)
     [r c]=size(x);
     t=0;
     if not(isempty(x))
     for i=1:r
         for j=1:c

             if abs(x(i,j))>5
                 t(i,j)=x(i,j);
             else
                 t(i,j)=0;
             end
             end
         end
     end
     y=real(double(t));

Answer 1

[Guillaume]

As per James' comment it would probably be more useful to fix your thresholding code so it works on empty cells as well.

Also note that in your example, the any is useless and fx could just be @isempty.

To answer your immediate question, you will have to use a loop:

for idx = 1:numel(c)
    c{idx}(cellfun(@isempty, c{idx})) = {nan};
end

Or if you really want to use cellfun for the outer cell array, you will have to use a named function (in a file) as anonymous functions can't do assignment:

function c = empty2nan(c)      %named function
  c(cellfun(@isempty, c)) = {nan}
end
c = cellfun(@empty2nan, c, 'UniformOutput', false)  %cellfun for outer cell array

The explicit loop is simpler, imo.

[GEEVARGHESE TITUS]

Dear Guillaume, Thank you for the quick answer. The code is working for faster than the one i had previously stated.

c={[] [] [2 4 5] [] [2 3 4 6 7]}

But the code is not working for cell with empty cells in at sub levels, for example

c={{[] [1 2]} { [] [] [2 4 5]} {[] [2 3 4 6 7] []}}

Also the first function is returning the error "Error: The expression to the left of the equals sign is not a valid target for an assignment." I feel the problem is that we are not able to address the sub indexed cells.

[Stephen23]

>> c = {{[],[1,2]},{[],[],[2,4,5]},{[],[2,3,4,6,7],[]}};
>> for idx = 1:numel(c)
       c{idx}(cellfun(@isempty, c{idx})) = {nan};
end
>> c{:}
ans = 
    [NaN]    [1x2 double]
ans = 
    [NaN]    [NaN]    [1x3 double]
ans = 
    [NaN]    [1x5 double]    [NaN]

There was a missing parenthesis.

@Guillaume: I added the missing parenthesis to your answer as well.

[GEEVARGHESE TITUS]

Thank you, Stephen .. got it

[GEEVARGHESE TITUS]

I have mentioned a coarse threshold operation, to apply to the cell.

function y = sthresh(x)
     [r c]=size(x);
     t=0;
     if not(isempty(x))
     for i=1:r
         for j=1:c
             if abs(x(i,j))>5
                 t(i,j)=x(i,j);
             else
                 t(i,j)=0;
             end
             end
         end
     end
     y=real(double(t));

but it is not skipping the empty or NAN cell, instead assigns zero to it. I used the cellfun to apply the function to a cell. What can be done to rectify this error.

[Guillaume]

"it is not skipping the empty or NAN cell, instead assigns zero to it."

Of course, it does. You've coded your function to make it so. Your function

function y = sthresh(x)
   t = 0;    %<-----  set output to 0!
   if ~isempty(x)
      %do something
   end
   y = t;
end

When x is empty, you put 0 in y. The simple way to work around this problem is to do:

t = zeros(size(x));  %instead of 0.

Which would also fix two other problems with the loop: 1) your loop grew the t on each step of the loop. 2) your loop didn't work properly for arrays with more than 2 dimension.

But of course, the loop isn't needed at all. I also don't understand the purpose of the real(double(t)) conversion. Assuming it's not needed, the function could simply be:

function x = sthresh(x)
   x(abs(x) <= 5) = 0;
end

which works on any size matrices, including empty ones. It also leaves NaN alone. If you want to change NaN to 0:

x(isnan(x) | abs(x) <= 5) = 0;

[GEEVARGHESE TITUS]

Thank u Guillaume. The code worked like a charm.