histcounts on matrice row by row, how to get rid of the loop?

Hi All, I have got a simple Matrice let say M( double 10*100) I need histcounts of every row, so far i do:

MhistCountsRow=zeros(10,10)
for iLoop=1:10
 MhistCountsRow(iLoop,:)=histcounts(M(iLoop,:));
end

I hate loop, i think they look disguting, how can i get rid of this one? Cheers Medie

Answer 1

[Star Strider]

Interesting that histcounts will not do what hist does easily.

For an illustration, try this — no loops necessary:

x = randn(100,3);
[N,edges] = hist(x, 25, 1);

figure(1)
bar3(edges,N)
grid on

EDIT — Using a cell array with histcounts avoids the loop:

x = randn(3,100);
xc = mat2cell(x, ones(1,size(x,1)), size(x,2));                 % Split Matrix Into Cells By Row
[hcell,hedges] = cellfun(@(x) histcounts(x,25), xc, 'Uni',0);   % Do ‘histcounts’ On Each Column
hmtx = cell2mat(hcell);                                         % Recover Numeric Matrix From Cell Array
edges = cell2mat(hedges);                                       % Recover ‘edges’ For Each Histogram

[the cyclist]

If you take this approach, be aware use of hist() is "discouraged" by Mathworks, and might become deprecated in a future release.

[Mederic Mainson]

Hi Star Strider, Yeap I know, the function i use to used histc does it so easy, but it has been superseeded by histcounts, I am just trying to stay up to date...

[Star Strider]

Agreed, but so long as we still have it, we might as well use it.

It’s regrettable that such useful functionality is disappearing.

[Star Strider]

@Mederic Mainson — I thought about it further and decided to use the cell functions to first break the matrix into individual columns (in my code) of cell vectors and then do histcounts on each cell vector. I then converted them back into numeric arrays.

You may have to experiment with my code to get it to work with your data, but it will work.

[Mederic Mainson]

Hi Star strider, I like your cell approach, its tricky but it works and get rid of the loop :) Thanks

[Star Strider]

My pleasure!

[Guillaume]

It's arguable that cellfun get rid of the loop. cellfun is just a loop by another name. The equivalent is commonly called foreach in other languages.

More importantly, you need to be aware that cellfun may actually be slower than a loop, particularly the way it is used here. You have the additional cost of computing a cell array, plus the cost of an anonymous function call on each element of the cell array, plus the cost of converting two cells arrays to matrices.

On the other hand cellfun and co. have the benefit of making it clear you operate on a whole sequence and that the same operation applies to each element of the sequence. See functional programming.

Answer 2

[the cyclist]

According to the documentation,

"Data to distribute among bins, specified as a vector, matrix, or multidimensional array. If X is not a vector, then histcounts treats it as a single column vector, X(:)."

So, you are out of luck hoping to avoid a loop over several vectors.

Already, your bit of code is certainly much shorter than, say, the code for the histcounts function itself, which hides all the complexity of the histogram algorithm.

If you don't want to see the for loop, then I suggest you take a similar approach. Embed the for loop in a function of your own -- maybe call it histcountsMatrix, and then just call the function as a one-liner.

[Mederic Mainson]

Hi cyclist, Thanks for your answer, i could create function for sure, but i use to use histc and this would only take one line, so i think there probably is a way out there to repeat this behaviour...?