I tried to implement both of your algorithms in MATLAB, and their sampling effectiveness is very low.
Wij = 1, i,j = 1,2, ... 20, the weight matrix W should produce factorial(20)~ 2.4329e+18 unique permutations, but:
algorithm1: for N = 1e5 samples produce only ~ 47000 unique permutations
algorithm2: for N = 1e5 samples produce only ~ 25000 unique permutations
A0 was defined as diagonal identity matrix 20 x 20
Could you please check out sampling effectiveness of your original R code and let me know?
I tried to implement both of your algorithms in MATLAB, and their sampling effectiveness is very low. Wij = 1, i,j = 1,2, ... 20, the weight matrix W should produce factorial(20)~ 2.4329e+18 unique permutations, but:
algorithm1: for N = 1e5 samples produce only ~ 47000 unique permutations algorithm2: for N = 1e5 samples produce only ~ 25000 unique permutations
A0 was defined as diagonal identity matrix 20 x 20
Could you please check out sampling effectiveness of your original R code and let me know?