Some theorems related to almost sure convergence of orthogonal series (Q5935905)

The author proves two theorems on almost sure convergence of random sequences. The first of them reads as follows. Let \((a_n)_{n\geq 1}\) be a sequence of positive numbers. Let \(\varphi: [0,\infty)\to [1,\infty)\) be an increasing map such that \(x\to x\varphi^2(x)\) is convex and (a) \(\int^1_0 {d\varepsilon\over \varepsilon\varphi(\varepsilon^{- 2})}< \infty\); (b) there exists \(\alpha> 1\) such that \(\int^\infty_1 {\varphi(x^\alpha)\over x^\alpha} dx< \infty\); (c) \(\sum_n a^2_n\varphi^2(a^{-2}_n)< \infty\). Then any random sequence \((X_n)_{n\geq 1}\) that satisfies the increment condition \(\|X_n- X_m\|_2\leq (\sum^m_{k= n+1} a^2_k)^{1/2}\), \(n< m\), is almost surely convergent. Moreover, \({\mathbf E}\sup_{n,m\geq 1}|X_n- X_m|< \infty\). Applications are given to a.s. convergence of orthogonal series and a.s. convergence of generalized Riemann sums. Indications are also given how the results obtained apply to quasi-orthogonal systems and ergodic averages.