Given a vector, I would like to sample without replacement elements from it repeatedly.

Given a (m x 1) vector v , I would like to ,randomly without replacement, sample s elements from it. I know I can use randsample(v,s) if I were to do this once. However, I want to do this repeatedly without using a for loop (i.e vectorization) so that it is fast.

[Matt J]

However, I want to do this repeatedly without using a for loop

If you make s>1, you will receive more than one selected sample. It will give you s samples in a single call.

Answer 1

[Matt J]

m=10; n=15; s=3;
V=rand(m,n);    %hypothetical data

[~,ids] = sort( rand(size(V)) ,1);
ids=ids(1:s,:)+m*(0:n-1);

selection=V(ids)  %selects s elements from each column of V

[Cem Gormezano]

Thank you Matt, just wondering wouldnt sort make things slower ?

[Matt J]

Slower than the loop? Not on my machine,

m=10; n=1e6; s=3;
V=rand(m,n);


tic;
[~,ids] = sort( rand(size(V)) ,1);
ids=ids(1:s,:)+m*(0:n-1);


selection=V(ids);
toc; %Elapsed time is 0.306036 seconds.

tic
for i=1:n
   randsample(V(:,i),s); 
end
toc%Elapsed time is 4.263803 seconds.

[Cem Gormezano]

Not all heroes wear capes, thanks !!!

[Matt J]

You're quite welcome, but please Accept-click the answer if it resolved your issue.

Answer 2

[Bruno Luong]

v = 'a':'z' % your vector
n = 10; % number of "loop"
s = 3; % number of drawing without replacement

[~,ir] = maxk(rand(n,length(v)),s,2);
r = v(ir)