Common random number simulation as the optimal coordinatewise coupling for iterated random functions
Sabrina Sixta, Jeffrey S. Rosenthal
Abstract
In this paper, we extend the results of Glasserman and Yao (1992) to demonstrate that common random number (CRN) simulation applied to coordinatewise monotone iterated random functions (IRF) represents the optimal allocation of random numbers to minimize the expected distance of an n-step Markov chain (MC) in Rq. We introduce this concept as the optimal coordinatewise Wasserstein distance (CWD) coupling. We provide specific examples of IRFs where CRN is the optimal CWD coupling, as well as cases where it is not. For IRFs that are not coordinatewise monotone, we illustrate how CRN applied to equimeasurable transformations can result in even lower expected distances compared to CRN applied to the original IRF.
Create a lesson
Related papers
BEAGLE in Action: Phylogenetic Likelihoods for Very Large State Spaces
Benjamin D Redelings, April Wright, Karthik Gangavarapu et al.
Fast nonparametric estimation of sparse precision matrices with a priori sparsity structure
Caleb Derrickson, Alexis Montoison, Vishwas Rao et al.
bacontrees: An extensible framework for Bayesian Context Trees in R
Victor Freguglia, Thiago Paulichen
Interval-valued circular time series: A cylindrical score-driven approach with an application to wind direction
Thasmika Mohan, Najmeh Nakhaei Rad, Sphiwe Skhosana et al.
Subgradient Methods for Nonsmooth Sample Average Approximation
Xi Chen, Ouqi Fu, Jiyuan Tu et al.
Spectral Gap Inheritance for Metropolis-Hastings with Reversible Proposals
Minhyeok Kim, Vinayak Rao