For Loop Vectorization

[mstachowsky]

Hello there. I'm a beginner in matlab and have come up against a problem that, for the life of me, i can't vectorize or make run any more efficiently. the goal is this:

you have an 8 by 8 grid that is placed on top of a coordinate plane that goes from -100 to 100 in x and y. so each grid point is a 25x25 square.

You also have a two-D vector of x,y coordinates, and you want to search that vector and find out how many of its coordinates end up in each square. (this is the "coords" vector i'm pasting below). you then do a simple calculation for entropy, but that's part of the problem that i don't have much trouble with

so my question is: I'm using 2(!) nested for loops, and since your goal is to run this program for 1 million elements, it takes far too long to run. is there a way to vectorize (or at least make faster) this program without resorting to cutting out some of the times you run through the loops?

here's the code:

entropy = zeros(8,8);
for k=1:8
%now we figure out entropy
for h = 1:8
n=length(find(coords,1)<(-100 +(k*200/8)) & coords,1)>(-100 + (k-1)*200/8) & coords,2) < (-100 + (h*200/8)) & coords,2)>(-100 + (h-1)*200/8)));
if(n~=0)
entropy(k,h) = -(n/400)*(log(n/400));
else
entropy(k,h)=0;
end
end
end

and of course we can assume that all variables that aren't already there have been declared and are full. coords is a 400x2 matrix.

Thanks,

Mike

 

[MikeyP]

What you are looking for is a 2-dimensional histogram.

There are plenty on the matlab central file exchange:

http://www.mathworks.com/matlabcentral/fileexchange/loadCategory.do

I suugest you try a few and see which is the fastest.

M

--
Dr Michael F Platten