Operation without for loop

I have two matrix.MAtrix A(250000x4)and MAtrix B(250000x4)

I have a function which result a scalar C.

    C = myfunction(X,Y);

Now i want to pass each row of MAtrixA and MAtrixB to myfunction without using a for loop.

How to do it ?

Answer 1

[Andrei Bobrov]

Y = arrayfun(@(i1)myfunction(A(i1,:), B(i1,:)), 1:size(A,1))

[Walter Roberson]

That's what I was thinking.

Now... can it be done with bsxfun() ? ;-)

[Sean de Wolski]

Sure. Permute B so that it is [1x4x250000].

bsxfun(@myFunction,A,permute(B,[3 2 1]));

It's going to use a ton of memory and probably be slower though. Of course, it's still awesomer because it uses bsxfun( :-).

[Sukuchha]

Thanks all for the input, i used andrei method. It seems to be working but havent check its performance of against for loop.

Answer 2

[Walter Roberson]

You can make it look prettier using arrayfun(), but arrayfun() is just going to be doing the "for" loop internally.

[Sukuchha]

is arrayfun() is faster than for loop ? I tried with arrayfun but arrayfun in passing one element of matrix A and one element of matrix B, not the whole row of matrix A and matrix B .

fh = @(x,y)(myfunction(x, y));

[Y]=arrayfun(fh, A,B,'UniformOutput',false)

[Walter Roberson]

I would not expect arrayfun() to be faster than a "for" loop. Your initial question was, however, not about performance; it was about applying the operation without using a "for" loop.

[Sukuchha]

yeh, i wanted something which would be faster than for loop, otherwise i would have done with for loop :)

[Walter Roberson]

Often in such cases the best thing to do is rewrite the function so that it accepts arrays.

Note that "for" loops are not necessarily slow, especially in new versions. There is probably more overhead in calling myfunction all those times than there is in the "for" loop itself.