News

On the frequency of false alarm of CUSUM change detection algorithms for dependent data

ARGOMENTI: Seminari

SEMINARI DI CALCOLO DELLE PROBABILITA`
La Dott.ssa Cecilia Prosdocimi, del Dipartimento di Matematica Pura e Applicata dell'Università di Padova, il 10 febbraio alle 10:00 in Aula 2BC/30 terrà un seminario dal titolo "On the frequency of false alarm of CUSUM change detection algorithms for dependent data".

-Abstract
On-line change detection concerns the sequential detection of stochastic inhomogeneities in a set of data. The CUSUM (cumulative sum) is the most popular change detection algorithm and its properties, when the data form an independent sequence, are well understood. In this talk we prove that the CUSUM statistic is L-mixing when the data form an L-mixing process. The class of L-mixing processes has proved to be very useful in system identification and its main properties, including a Law of Large Numbers, will be reviewed. As a by product of the main result we will provide an estimator of the false alarm frequency of the CUSUM for L-mixing data.

This is a joint work with Laszlo Gerencser, SZTAKI Institute, Hungarian Academy of Science, Budapest.

Rif. int. P. Dai Pra

NEWS: New Second Level Degree in Data Science - Second cycle degree - a. y. 2017/18 X