Volume 22, 2018
|Page(s)||129 - 162|
|Published online||16 November 2018|
Uniform convergence of penalized time-inhomogeneous Markov processes
IECL (UMR CNRS 7502), Université de Lorraine,
Site de Nancy, B.P. 70239,
Vandœuvre-lès-Nancy Cedex, France.
2 Inria, TOSCA Team, Villers-lès-Nancy 54600, France.
* Corresponding author: Nicolas.Champagnat@inria.fr
Accepted: 12 December 2017
We provide a general criterion ensuring the exponential contraction of Feynman–Kac semi-groups of penalized processes. This criterion applies to time-inhomogeneous Markov processes with absorption and killing through penalization. We also give the asymptotic behavior of the expected penalization and provide results of convergence in total variation of the process penalized up to infinite time. For exponential convergence of penalized semi-groups with bounded penalization, a converse result is obtained, showing that our criterion is sharp in this case. Several cases are studied: we first show how our criterion can be simply checked for processes with bounded penalization, and we then study in detail more delicate examples, including one-dimensional diffusion processes conditioned not to hit 0 and penalized birth and death processes evolving in a quenched random environment.
Mathematics Subject Classification: 60B10 / 60F99 / 60J57 / 37A25
Key words: Feynman–Kac formula / time-inhomogeneous Markov processes / penalized processes / one-dimensional diffusions with absorption / birth and death processes in random environment with killing / asymptotic stability / uniform exponential mixing / Dobrushin’s ergodic coefficient.
© EDP Sciences, SMAI 2018
Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.