Skip to main content
. 2022 Jan 8;24(1):99. doi: 10.3390/e24010099
Algorithm 1 Solve globally optimal ZOSC with channel gains from finite alphabet M
0. Initialize the marginal distributions PX1∈R+|M|×1,PX2∈R+|M|×1,PY∈R+|M|×1 and define P:=[PX1T,PX2T,PYT]T.
1. Construct the |M|-ary expansion matrix E, where each row of E is a tuple (x1,x2,y)∈M3 (the 1st row is (0,0,0) and the proceedings follow an increasing order with respect to the |M|-ary expansion).
2. Construct A∈[0,1]3|M|×|M|3, where Ai,j=0, if by the j-th row of E, x1+x2≤y and pj is not used in calculating the marginal probability Pi, where Pi is the i-row of P, i=1,⋯,3|M|, j=1,⋯,|M|3. Define L=3|M|.
3. Reorder A and E as A˜ and E˜, respectively, such that {12lng(Ei˜)}i∈L are in an increasing manner.
repeat
  4. Update A˜: set L˜−L columns of A˜ as a zero vectors, where the indices of those columns correspond to the rows of E˜ having the smallest rates.
  5. Solve A˜p˜˜˜=P˜, 0<p˜˜˜i≤1, ∑pi=1, i∈L˜, by CVX.
  6. Set L = L + 1.
untilA˜p˜˜˜=P˜ is feasible,
7. ZOSC = min(x1,x2,y)∈F12[log2(1+x1+x2)−log2(1+y)], where
F:={Ej|ifpj>0,j=1⋯,|M|3}.