%A = [1;5;2]
A = [5;5;2]
B = NaN(3,6);
% new code:
idx = A==max(A);
num = nnz(idx);
A(idx) = max(A)-num+randperm(num,num);
% original code:
idy = A==max(A);
B(idy,1:3) = A
Note that this works for one or more maximum values. The code does not randomly pick new values from 1:5 as you ask, but instead uses randperm on the set of values [max,max-1,...]. This is because if the values are randomly picked then there is no guarantee that the new values will not cause multiple new max values, eg:
>> A = [5;5;4];
>> new = [4,1]; % random between 1:5
>> A(A==max(A)) = new
A =
4
1
4
and thus you get back to the same problem as you had at the start. For this reason I used randperm.
