Gradient estimation for smooth stopping criteria

Bernd Heidergott; Yijie Peng

doi:10.1017/apr.2022.7

Gradient estimation for smooth stopping criteria

Bernd Heidergott^*, Yijie Peng

^*Corresponding author for this work

Research output: Contribution to Journal › Article › Academic › peer-review

46 Downloads (Pure)

Abstract

We establish sufficient conditions for differentiability of the expected cost collected over a discrete-Time Markov chain until it enters a given set. The parameter with respect to which differentiability is analysed may simultaneously affect the Markov chain and the set defining the stopping criterion. The general statements on differentiability lead to unbiased gradient estimators.

Original language	English
Pages (from-to)	29-55
Number of pages	27
Journal	Advances in Applied Probability
Volume	55
Issue number	1
Early online date	15 Jun 2022
DOIs	https://doi.org/10.1017/apr.2022.7
Publication status	Published - Mar 2023

Bibliographical note

Publisher Copyright:
© The Author(s), 2022. Published by Cambridge University Press on behalf of Applied Probability Trust.

Keywords

gradient estimation
Monte Carlo simulation
Sensitivity analysis

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

Access to Document

10.1017/apr.2022.7

Gradient-estimation-for-smooth-stopping-criteriaFinal published version, 430 KBLicence: Article 25fa Dutch Copyright Act

Persistent URL (handle)

Persistent link

Cite this

@article{1c954541138a46e7b58ceabbcd09723a,

title = "Gradient estimation for smooth stopping criteria",

abstract = "We establish sufficient conditions for differentiability of the expected cost collected over a discrete-Time Markov chain until it enters a given set. The parameter with respect to which differentiability is analysed may simultaneously affect the Markov chain and the set defining the stopping criterion. The general statements on differentiability lead to unbiased gradient estimators.",

keywords = "gradient estimation, Monte Carlo simulation, Sensitivity analysis",

author = "Bernd Heidergott and Yijie Peng",

note = "Publisher Copyright: {\textcopyright} The Author(s), 2022. Published by Cambridge University Press on behalf of Applied Probability Trust.",

year = "2023",

month = mar,

doi = "10.1017/apr.2022.7",

language = "English",

volume = "55",

pages = "29--55",

journal = "Advances in Applied Probability",

issn = "0001-8678",

publisher = "Cambridge University Press",

number = "1",

}

TY - JOUR

T1 - Gradient estimation for smooth stopping criteria

AU - Heidergott, Bernd

AU - Peng, Yijie

N1 - Publisher Copyright: © The Author(s), 2022. Published by Cambridge University Press on behalf of Applied Probability Trust.

PY - 2023/3

Y1 - 2023/3

N2 - We establish sufficient conditions for differentiability of the expected cost collected over a discrete-Time Markov chain until it enters a given set. The parameter with respect to which differentiability is analysed may simultaneously affect the Markov chain and the set defining the stopping criterion. The general statements on differentiability lead to unbiased gradient estimators.

AB - We establish sufficient conditions for differentiability of the expected cost collected over a discrete-Time Markov chain until it enters a given set. The parameter with respect to which differentiability is analysed may simultaneously affect the Markov chain and the set defining the stopping criterion. The general statements on differentiability lead to unbiased gradient estimators.

KW - gradient estimation

KW - Monte Carlo simulation

KW - Sensitivity analysis

UR - http://www.scopus.com/inward/record.url?scp=85148488529&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85148488529&partnerID=8YFLogxK

U2 - 10.1017/apr.2022.7

DO - 10.1017/apr.2022.7

M3 - Article

AN - SCOPUS:85148488529

SN - 0001-8678

VL - 55

SP - 29

EP - 55

JO - Advances in Applied Probability

JF - Advances in Applied Probability

IS - 1

ER -

Gradient estimation for smooth stopping criteria

Abstract

Bibliographical note

Keywords

UN SDGs

Access to Document

Persistent URL (handle)

Other files and links

Fingerprint

Cite this