Date:
Tue, 14/04/202614:00-15:00
Location:
Ross 70
Title: Thresholds for Poisson Limits in Symbolic Dynamics
Speaker: Nicolò Paviato (Weizmann Institute)
Abstract:
About twenty-five years ago, Peres and Weiss generalised the classical Poisson limit theorem for appearances of words of increasing length in a sequence x. They showed that the theorem holds for almost every x with respect to the infinite uniform product measure. A natural question is whether this Poisson behaviour persists when the sequence is sampled according to a different product measure. In our first result, we consider non-stationary product measures and show that there exists a quantitative threshold above which the Poisson limit theorem holds for almost every x, while below this threshold it may fail. In contrast, our second result shows that for a biased infinite product measure (a non-fair coin) the limiting behaviour is almost surely non-Poisson. This shows that the Poisson regime is specific to the equiprobable case and to small deviations from it.
This talk is based on works with Mike Hochman and Jon V. Kogan.
Speaker: Nicolò Paviato (Weizmann Institute)
Abstract:
About twenty-five years ago, Peres and Weiss generalised the classical Poisson limit theorem for appearances of words of increasing length in a sequence x. They showed that the theorem holds for almost every x with respect to the infinite uniform product measure. A natural question is whether this Poisson behaviour persists when the sequence is sampled according to a different product measure. In our first result, we consider non-stationary product measures and show that there exists a quantitative threshold above which the Poisson limit theorem holds for almost every x, while below this threshold it may fail. In contrast, our second result shows that for a biased infinite product measure (a non-fair coin) the limiting behaviour is almost surely non-Poisson. This shows that the Poisson regime is specific to the equiprobable case and to small deviations from it.
This talk is based on works with Mike Hochman and Jon V. Kogan.
