Replicate the vector so it has more elements than you want. Then use the two input form of randperm to select the desired number of elements from that replicated vector. Alternately if you're okay with selecting with more duplicate elements than would be in the replicated vector, use randi on the original vector.
x = 1:5;
n = 12; % Ask for 12 elements from x
replicatedX = repmat(x, 1, ceil(n/numel(x)))
This randperm approach would select no more than 3 instances of each of the elements from 1 to 5, since the ceiling of 12/5 is 4.
y1 = replicatedX(randperm(numel(replicatedX), n))
histogram(y1) % No bar is higher than 3
This randi approach could generate a vector with more than 3 of any elements but doesn't require the replicated vector.
y2 = x(randi(numel(x), 1, n))
figure
histogram(y2) % Could have bins higher than 3
