A proof of the Shepp-Olkin entropy concavity conjecture (Q2405168)

In [in: Contributions to probability, Collect. Pap. dedic. E. Lukacs, 201--206 (1981; Zbl 0534.60020)], \textit{L. A. Shepp} and \textit{I. Olkin} formulated the following conjecture: the Shannon entropy of a sum of independent heterogeneous Bernoulli random variables (that is a Poisson binomial random variable) is a concave function in the parameters of respective Bernoulli distributions. They have shown that this property is true for some particular cases. In the present paper, this subtle conjecture is fully proved. For the proof, the authors use a refined technique. Note that, the same authors have previously solved the `monotone' case of the Shepp-Olkin entropy concavity conjecture [\textit{E. Hillion} and \textit{O. Johnson}, Ann. Probab. 44, No. 1, 276--306 (2016; Zbl 1348.60139)]. The result on Shannon entropy is then extended to the entropy in the meanings of Rényi and of Tsallis.