Estimating random polynomials by means of metric entropy methods (Q4504497)

The author considers random polynomials of the form \[ Z_N(t)= \sum^N_{k=1} \theta_k\{X_k\cos 2\pi p_kt +Y_k\sin 2\pi p_kt\}, \] where \((\theta_k)\) is a sequence of real numbers, \((p_k)\) is a strictly increasing sequence of integers greater than 1, and \((X_k)\), \((Y_k)\) are two sequences of random variables defined on the same probability space \((\Omega, {\mathcal A},P)\). Under an extra assumption on the increments of the process \((Z_N)\), the author shows that a universal constant \(C\) exists, such that for all \(N\geq 1\), \[ \Bigl \|\sup_{0\geq t\geq 1}\bigl|Z_N(t) \bigr|\Bigr \|_G\leq C(\log p_N)^{1/2} \left(\sum^N_{k=1} \theta^2_k \right)^{1/2}, \] where \(\|\;\|_G\) denotes the gauge norm on \(L^G(P)\), the space of all measurable functions \(f\) satisfying \(E(\exp(a^2f^2) -1)<\infty\) for some real \(0<a <\infty\). Several applications of the above result are given, one of which is the well-known Salem-Zygmund's bound for random polynomials as well as a sufficient condition for the uniform convergence of random Fourier series.