Let $H$ be a Hilbert space, and $X_n$, $ n\in \mathbb {Z}$, be a sequence of independent Bernoulli random variables $P(X_n = \pm 1) = \frac 12$. Is there a characterization of the sequences $a_n$, $ n\in \mathbb{Z}$, in $H$ such that the series $$ \sum_n a_n X_n e^{int}, \quad t \in [0,2\pi], $$ is almost surely uniformly convergent in $H$?

  • 3
    $\begingroup$ Did you look in J.-P. Kahane’s book, Some Random series of functions. I believe this will contain answers to your question. $\endgroup$ – Anthony Quas Feb 23 at 7:40
  • $\begingroup$ As far as I can tell, there's nothing in Kahane's book about that, not even a characterization of scalar sequences. $\endgroup$ – Dang Zheng Feb 27 at 8:01
  • 2
    $\begingroup$ @DangZheng That's partially wrong: there is no full characterization (for the simple reason that it is still unknown) but there is a whole chapter on when the sum of a random Fourier series with scalar coefficients is almost surely continuous and the continuity is actually derived from the uniform convergence. $\endgroup$ – fedja Feb 27 at 16:30

There is actually a full characterization for scalar sequences in the work of Marcus and Pisier (1981) via entropy integrals. The vector-valued case is an application of a deep result on Bernoulli processes (Bednorz-Latala, Annals of Mathematics 2013).


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.