A Gaussian stationary function (GSF) is a random f: R --> R whose
distribution is shift-invariant and all its finite marginals have
centered multi-normal distribution. It is a simple and popular model
for noise, for which the mean number of zeroes was computed already
in the 1940's by Kac and Rice. However, it is far more complicated
to estimate the probability of a significant deficiency or abundance
in the number of zeroes in a long interval (compared to the expectation).
We do so for a specific family of GSFs with additional smoothness and absolutely
summable correlations, using tools from real and complex analysis.
Joint work with R. Basu, A. Dembo and O. Zeitouni
arXiv:1709.06760.