Code optimization by avoiding loops

I have the following Data and Variables

data = 1000x2 double array containing x and y coordinates
B = 1000x1 integer array containing category values from 1 to j inclusive
A = need to create this array which contains values that are the mean of all the points for each category (x and y values to be averaged separately)

I have the following code which works as desired. I am wondering whether there is a better way of doing it without the loop,

for i = 1:j
    A(i,:) = mean(data(B==i,:));
end

Thank you.

Answer 1

[Matt J]

B=B(:);
N=accumaray(B,1);
A(:,1)=accumaray(B,data(:,1))./N;
A(:,2)=accumaray(B,data(:,2))./N;

[Matt J]

You could also use accumarray(...,@mean) but in past Matlab versions, at least, that has been found to be slower.

[Alex]

I'm new to matlab and not familiar with that expression? Can you please suggest the full code out of interest? Your previous code works perfectly btw. Thanks a bunch.

[Matt J]

Can you please suggest the full code out of interest?

 A(:,1)=accumaray(B(:),data(:,1),[],@mean);
 A(:,2)=accumaray(B(:),data(:,2),[],@mean);

I'm new to matlab and not familiar with that expression?

Since you're new, it will benefit you to know that you can bring up usage documentation on any command, e.g.,

>> doc accumarray

[Alex]

Thanks. Can the same be done when plotting the values? My current code is,

    for i = 1:j
        C = data(B==i,:);
        plot(C(:,1),C(:,2));
    end

[Matt J]

Once you read "doc accumarray", you will know ;)

[Matt J]

A fancy trick you might be able to do is

args=[accumarray(B(:),data(:,1),[],@(q) {q}),...
      accumarray(B(:),data(:,2),[],@(q) {q})].';

plot(args{:})