Vectorization of loop with find

% I have always wondered how I might be able to vectorize this:

l=0:0.1:1;

a = zeros(length(l),1);

x = rand(1000000,1); % some large array

for i=1:length(l)

    a(i) = length(find(x>l(i)));

end

% In the end, basically I just need a count of how many elements of x exceed each value of l

Thank you for your help.

Answer 1

[Fangjun Jiang]

use hist()

The example below shows how it can be used. The result is the same as a although it is in a slightly different format initially. Note sometimes the numbers may be off by 1. I think it's due to border-line number like 0.1, 0.2, etc. I think you'll get the idea.

x = rand(1000,1); % some large array
b=hist(x);
c=b(end:-1:1);
d=cumsum([0,c]);
l=0:0.1:1;
a = zeros(length(l),1);
for i=1:length(l)
    a(i) = length(find(x>l(i)));
end

RevA=a(end:-1:1);

all(d'==RevA)

Answer 2

[Walter Roberson]

Use histc() instead of hist!

a = flipud(cumsum(flipud(histc(x(:), [l*(1+eps),inf], 1))));

The additional argument of 1 to histc fixes the obscure behavior that occurs if x happens to be a scalar.

Answer 3

[Nirmal]

Great suggestions. Thanks!

Answer 4

[Andrei Bobrov]

variant

sum(bsxfun(@gt,x(:),0:.1:1))