Apply function to each column of matrix for all columns without for loop

Hi all,

I have a function "Func(X)" which operations on n-by-1 single column matrix "X" and outputs a scalar quantity. I also have a n-by-m matrix "A" that I would like to apply "Func" for all m columns in "A" and obtain 1-by-m matrix, say B, as a result without using for loop or any iterative definitions. What is the best way to proceed with this?

[madhan ravi]

upload your code and datas

[Jeongseop Lee]

Here is it, as an example. The actual function is complicated and there is a little room for modification but for illustration I will just define a much simpler operation that is essentially equivalent.

% example for "A"
A = [1 2 3 4;5 6 7 8;9 10 11 12];

% definition for "Func"
function Result = Func(X)
[m,n] = size(X);
if n>1
    error
else 
    Result = sum(X)
end

[Paolo]

If B is the output matrix, what do you want B to be for that A?

[Jeongseop Lee]

The actual calculation of B is much more complicated that takes cascades of multiple functions and B is an optimization fitness product for given population A in neural network. But all this detail is highly irrelevant for my question here. I just want a simple, elegant way that is non-iterative definition that applies a function to each column of a given matrix for all columns without iterative definition.

Answer 1

[Matt J]

There is no general way that is more efficient than a for-loop, but if you are just trying to hide the loop, then you can do things like this:

 B=cell2mat( arrayfun(@(i) Func(A(:,i)) , 1:size(A,2), 'uni',0) );

or

 B = splitapply(@Func,A,1:size(A,2));

Answer 2

[Greg Heath]

The MATLAB CONVENTION is that functions operate on matrix columns.

Therefore, it is only with user-defined functions that operations on rows occur.

Hope this helps.

Thank you for formally accepting my answer

Greg