Scalar max of cell array with structure

I have a cell array "myCell" with 2 dimensions and each cell has a 2D matrix and I'd like to obtain scalar value of the maximum for a given structure among all the cell dimensions. I assume bsxfun can do this but my goal doesn't match they way that I normally see bsxfun get used.

Currently I get the max (of the abs) looping through each cell dimension, grabbing the max for all elements in that cell's structure I need, assign the (scalar) max to elements in a matirx which shares the cell array dimensions, then tak the max of the matrix with (:) as shown below.

peak_FcRn(noFc,noRa)=0; % each cell array has  structure, whose maxes I want to compare

for iFc=1:noFc
    for iRa=1:noRa
        % get the maximum absolute value of the current cell array's structure 
        % called noise (myCell{iRa,iFc}.noise--2 or 3d array)
        peak_FcRn(iFc,iRa) = max(abs(myCell{iRa,iFc}.noise(:))); % a scalar value
    end
end

% the
peak = max(peak_FcRn(:)); %

This gets the right answer but I would really like a one- or two-liner that can get the max of the cell's structure elements--and I have a feeling bsxfun can do it!

In the laziest way I would type

peak = max(abs(myCell{:,:}.noise(:))) % or perhaps
peak = max(abs(myCell{:}.noise(:))) 

Thanks for your input,

Michael B.

[darova]

Try cellfun

[Michael Bowles]

Thanks for your input. I got it to work with a cute two-liner!

peakPerCELL = cellfun(@(x) max(abs(x.noise(:))),myCell);
peak = max(peakPerPADS(:));

Cheers!

Michael B.

Answer 1

[James Tursa]

If you reshape the input first, a one-liner version of your code:

peak = max(cellfun(@(x) max(abs(x.noise(:))),myCell(:)));

[Michael Bowles]

Thanks that's even neater! Can you explain why taking the max outside cellfun ignores the dimensions (i.e. doesn't need an associated "(:)"? I realize I could use

peak = max(cellfun(@(x) max(abs(x.noise(:))),myCell(1,:)));

to grab along a chosen dimension, but why is the default to treat the output of cellfun (a whos tells me it's a double) like a 1D vector?

Thanks!

[James Tursa]

The myCell(:) reshapes myCell into a column vector and passes that into cellfun( ), which in turn creates the output with the same shape as the input ... a column vector. Since the output is now already a column vector, no need to reshape it again before taking max( ).

[Michael Bowles]

I understand and realize my question is silly now! I confused my self before posting my second question because I thought I hadn't changed the last cellfun input with (:).

Thanks for your help! I learned great stuff today!