EXPONENTIAL INEQUALITIES FOR SUPREMA OF PROCESSES WITH STOCHASTIC NORMALIZATION
Abstract
We prove a deviation bound for the supremum of a normalized process derived from a real-valued random process $(Y_t)_{t \in S}$. Assuming the increments $Y_t - Y_s$ satisfy Bernstein deviation bounds with random fluctuation terms makes it possible to use a stochastic normalization term. This improves on classical versions where the control of the fluctuation is deterministic.
Fichier principal
arxivconcentration.pdf (289.5 Ko)
Télécharger le fichier
Supremawithstochasticnormalization.pdf (289.48 Ko)
Télécharger le fichier
| Origin | Files produced by the author(s) |
|---|
| Origin | Files produced by the author(s) |
|---|
