Discrete entropies of orthogonal polynomials (Q836085): Difference between revisions

The authors introduce the discrete entropy of orthonormal polynomials defined as the Shannon entropy of the probability distribution constructed as follows. For a nontrivial Borel measure on the real line, they consider the corresponding orthonormal polynomial of degree \(n\), \(p_n\), whose zeros are \( \lambda_j^{(n)},\; j=1,\dots, n\). Then for each \(j=1,\dots, n,\) \[ \overrightarrow{\Psi}_j^2=(\Psi_{1j}^2,\dots,\Psi_{nj}^2)\;\; \text{with}\;\; \Psi_{ij}^2= p_{i-1}^2(\lambda_j^{(n)})\left(\sum _{k=0}^{n-1}p_k^2(\lambda_j^{(n)})\right)^{-1}, \] defines a discrete probability distribution. The Shannon entropy of the sequence \(\{p_n\}\) is defined as \(\mathcal{S}_{n,j}=- \sum _{i=1}^n \Psi_{ij}^2 \log (\Psi_{ij}^2).\) Since there are no known results for the Shannon entropy of the orthogonal polynomials' related distributions, it is important to point out that the authors compute explicitly the discrete entropy \(\mathcal{S}_{n,j}\) corresponding to the Chebyshev orthonormal polynomials of the first and second kinds and they also find the first two terms of the asymptotic expansion of \(\mathcal{S}_{n,j}\) for fixed \(j\) and large \(n\). The formulas exhibit nice connections with relevant objects from number theory. Finally, they present some results of numerical evaluation of the entropy for several orthogonal polynomials.