How to speed up find and unique

I did the profiling where unique and find take the most of time.

I use find as I want to know the row and column index of the 1 elements in a. For unique, I only need the index.

a=randi([0 1], 20000,8736);
temp=randi(8737, 4000,1000);

for i=1000
[aa0,aa00]=find(a(:,temp(:,i))==1);
[~,aa001,~]=unique(aa00);
end

May I know how to speed up? Thank you.

[Walter Roberson]

I think your aa001 should be equivalent to

aa001 = find(diff([0; aa0]));

which is an abbreviation for

aa001 = [1; find(diff(aa0) ~= 0)];

which is looking for the places where aa0 changes.

Potentially faster would be

aa001 = [1; find(aa0(1:end-1) ~= aa0(2:end))];

but with the two intermediate arrays you would really have to do a timing test to be sure which variation was faster.

[Chaoyang Jiang]

Thank you for your answer. The output of aa001 = find(diff([0; aa0])) is the unique value, while in my code, [~,aa001,~]=unique(aa00) the output is the unqiue value index. They are not same.