Time Averages of Stochastic Processes: a Martingale Approach
Pepin, Bob
arXiv, 1810.10945 / Harvested from arXiv
This work shows how exponential concentration inequalities for time averages of stochastic processes over a finite time interval can be obtained from a martingale representation formula. The approach relies on mixing properties of the underlying process, applies to a wide range of initial conditions and makes no assumptions on stationarity or time-homogeneity. A direct method is presented for diffusion processes and discrete-time Markov processes. For general square-integrable processes the constants in the concentration inequalities can be expressed in terms of the quadratic variation of a family of auxiliary martingale. For continuous-time Markov processes they admit a natural expression in terms of the squared field operator applied to the semigroup. The paper concludes with two examples: the squared Ornstein-Uhlenbeck process and the $M/M/\infty$ queue.
Publié le : 2018-10-25
Classification:  Mathematics - Probability,  60G17
@article{1810.10945,
     author = {Pepin, Bob},
     title = {Time Averages of Stochastic Processes: a Martingale Approach},
     journal = {arXiv},
     volume = {2018},
     number = {0},
     year = {2018},
     language = {en},
     url = {http://dml.mathdoc.fr/item/1810.10945}
}
Pepin, Bob. Time Averages of Stochastic Processes: a Martingale Approach. arXiv, Tome 2018 (2018) no. 0, . http://gdmltest.u-ga.fr/item/1810.10945/