Treffer: Perturbation analysis via coupling

Title:
Perturbation analysis via coupling
Authors:
Source:
IFAC Proceedings Volumes. 32:4753-4758
Publisher Information:
Elsevier BV, 1998.
Publication Year:
1998
Document Type:
Fachzeitschrift Article
File Description:
application/xml
Language:
English
ISSN:
1474-6670
DOI:
10.1016/s1474-6670(17)56810-x
DOI:
10.1117/12.319338
DOI:
10.1109/9.847099
Rights:
Elsevier TDM
IEEE Copyright
Accession Number:
edsair.doi.dedup.....2152b2e200fe0cfdba80f9fd7aa10c20
Database:
OpenAIRE

Weitere Informationen

The problem of gradient estimation consists in obtaining an estimate of \(J'(\theta)\) from a sample of \(X_\theta\), where \(X_\theta\) is a parameterized random variable, \(\theta\in (a,b)\subset \mathbb{R}\), \(J(\theta)= E[L(X_\theta)]\), \(L(x): \mathbb{R}\to\mathbb{R}\). A few general gradient estimators are derived using perturbation analysis via coupling. Specifically, the coupling method is applied to gradient estimation of Markov chains.