efficient use of find and ismember

Hello,

I have 2-D array (~250000 x 2) that has repeated rows. I need to find the indices of the repeated rows and take a mean over those rows with another variable. The code I have below works fine, but it is VERY slow. In this example the variable 'latlon' is the 2-D array (1st column is latitude and 2nd column is longitude).

uniq = unique([latlon],'rows');

for i = 1:length(uniq)
     pairs = find(ismember(latlon,uniq(i,:),'rows'));
     %use pairs indices to grab data from another 1-D variable
     newdata(i) = mean(SLP(pairs))
end

This currently takes ~4hrs to run for the 250,000 row array. Can anyone help me improve the efficiency here?

Thanks,

Dan

Answer 1

[dpb]

The "deadahead" solution (w/o accumarray, I'm to brain-tired this evening to write that off the cuff... :) )

[~,ia,ic]=unique(latlon,'rows','stable');
new=zeros(size(ia));
for i=1:length(ia)
  new(i) = mean(SLP(ia(i)==ic));
end

I think I got that right...