# How to vectorize a find

8 views (last 30 days)
Matlab2010 on 11 Dec 2012
x = zeros(1000,1);
x(76) = 1;
x(100) = 1;
x(200) = 1;
I can do:
first = find(x ==1, 1, 'first'); %the answer =76
but this is slow. I would like to speed this up. How? thanks
Matlab2010 on 15 Jan 2013
just going to put up some code which might help others looking for v fast finds in future:
Method 3 is incredibly fast (credit to Matt J!)
%METHOD 1
lastTimeValue_SLOW = arrayfun(@(x,y) find(data(x:end,2) <= y, 1, 'last'), trades, tradeTimePlusDt); %works but v slow!
%METHOD 2
for z = 1 : length(trades)
for kk = 1:length(p)
lastTimeValue(z) = kk-1;
break;
end
end
end
%METHOD 3
for z = 1 : length(trades)
temp = cumsum(I(1:z));
lastTimeValue_TEST(z) = temp(end);
end

Matt Fig on 11 Dec 2012
Edited: Matt Fig on 11 Dec 2012
The last statement you show is vectorized, so the question is ill-posed. I think what you are asking for is a way to speed up the process.
Also, just how fast do you think it should be??
x = zeros(1000,1);
x(76) = 1;
x(100) = 1;
x(200) = 1;
tic,first = find(x ==1, 1, 'first');toc
Elapsed time is 0.000023 seconds. % That is slow???
.
.
.
EDIT
I did find a way to do it faster, but it will only work if you have a binary matrix as in your example. You can avoid the comparison as it is not necessary.
tic,first = find(x);first = first(1);toc
Elapsed time is 0.000012 seconds.
Matt Fig on 11 Dec 2012
Also, see my edit above for a faster method if you have a binary array.

Sean de Wolski on 11 Dec 2012
If you only have zeros and ones and you are positive there is atleast one one, then you can use the second output from max().
[~,first] = max(x);
I don't know if this will be faster or not.

Mark Whirdy on 11 Dec 2012
Edited: Mark Whirdy on 11 Dec 2012
a temporary vector of row numbers, then use your vector-of-interest & a logicsl statement to index this row-number vector
a =[1:size(x,1)]'; % row numbers
b=a(x==1); % logical indexing the populated rows
b(1) % first instance