On the constant in Meńshov-Rademacher inequality (Q2491568): Difference between revisions

The Menshov-Rademacher inequality deals with the estimation of \[ D_n:=\sup E\left[\max_{1\leq k\leq n}\left(\sum^k_{\ell=1}\alpha_\ell \varphi_\ell \right)^2 \right], \] where the supremum is extended over all probability spaces \((\Omega, {\mathcal F},P)\), all real orthonormal systems \((\varphi_1,\dots, \varphi_n)\) on them, and all real coefficient collections \((\alpha_1,\dots, \alpha_n)\) with \(\sum^n_{\ell_1}\alpha_\ell^2=1\). Among others, it is proved that \[ D_2=\frac 43\quad\text{and}\quad \limsup_{n\to\infty}\frac{D_n}{\log^2_2n}\leq\frac 14. \] Reviewer's remark: The proof of the Menshov-Rademacher inequality by making use of a recurrence relation was also made by the reviewer [Z. Wahrscheinlichkeitstheor. Verw. Geb. 35, 299--314 (1976; Zbl 0314.60023)].