Maximum Entropy of Sums of Independent Ternary Random Variables
This provides a theoretical result for information theory, settling the ternary case of a classical entropy maximization problem.
The paper solves the problem of maximizing Shannon entropy for sums of independent ternary random variables, proving that the maximum is achieved by a specific distribution involving uniform variables on {0,2} and a weighted ternary variable. This extends the Shepp-Olkin-Mateev theorem to ternary alphabets.
The classical problem of maximizing the Shannon entropy of a sum of independent random variables supported on a finite alphabet is considered and settled in the ternary case. Namely, the following theorem is established: if \(X_1,\ldots,X_n\) are independent random variables taking values in \(\{0,1,2\}\), then the entropy of \(S_n=X_1+\cdots+X_n\) is maximized when \(X_1,\ldots,X_{n-1}\) are uniform on \(\{0,2\}\) and the probability mass function of \(X_n\) is given by \(\Prob(X_n=0) = \Prob(X_n=2) = w/2\), \(\Prob(X_n=1) = 1-w\), where \(w = \big(1 + 2^{-H(B_n)+H(B_{n-1})}\big)^{-1}\) and \(B_m\sim \Bin(m,1/2)\). The statement can be seen as an extension to ternary alphabets of the Shepp--Olkin--Mateev theorem. The proof uses the Hermite--Biehler theorem, Newton's inequalities, and Yu's maximum-entropy theorem for ultra-log-concave distributions.