Central Limit Theorems for Stochastic Optimization Algorithms Using Infinitesimal Perturbation Analysis

نویسندگان

  • Qian-Yu Tang
  • Han-Fu Chen
چکیده

Central limit theorems are obtained for the \perturbation analysis Robbins-Monro single run" (PARMSR) optimization algorithm, with updates either after every L customers or after every busy period, in the context of the optimization of a GI=GI=1 queue. The PARMSR algorithm is a stochastic approximation (SA) method for the optimization of innnite-horizon models. It is shown that the convergence rate and the asymptotic variance constant of the optimization algorithm, as a function of the total computing budget (i.e., total number of customers), are the same for both updating methods, and independent of L, provided that the step sizes of SA are chosen in the (asymptotically) optimal way for each method. R esum e. On obtient des th eor emes de la limite centrale pour la solution de l'algorithme \perturbation analysis Robbins-Monro single run" (PARMSR) pour l'optimisation d'une le d'attente GI=GI=1, avec mise a jour apr es chaque L clients ou apr es chaque cycle regeneratif. L'algorithme PARMSR est une m ethode d'approximation stochastique (AS) pour l'optimisation de mod eles sur horizon innni. Nos r esultats impliquent que le taux de convergence et la constante de variance asymptotique de l'algorithme d'optimisation, en fonction du budget total de calcul (i.e., le nombre total de clients), sont les m^ emes pour les deux m ethodes de mise a jour, et ind ependants de L, pourvu que les longueurs de pas de l'AS soient choisis de mani ere optimale pour chaque m ethode. We thank the two anonymous referees, whose comments helped improving the presentation of the results in paper.

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

منابع مشابه

Asymptotic E ciency of Perturbation Analysis Based Stochastic Approximation with Averaging

Central limit theorems are obtained for the PARMSR (perturbation analysis Robbins-Monro single run) algorithm with averaging, updated either after every regenerative cycle or after every xed-length observation period, for one-dependent regenerative processes. These stochastic approximation algorithms with averaging turn out to have identical limiting behavior, i.e., the same convergence rate an...

متن کامل

Convergence of a Stochastic Approximation Algorithm for the GI/G/1 Queue Using Infinitesimal Perturbation Analysis

Discrete-event systems to which the technique of infinitesimal perturbation analysis (IPA) is applicable are natural candidates for optimization via a Robbins-Monro type stochastic approximation algorithm. We establish a simple framework for single-run optimization of systems with regenerative structure. The main idea is to convert the original problem into one in which unbiased estimators can ...

متن کامل

Perturbation Analysis and Optimization of Stochastic Hybrid Systems

We present a general framework for carrying out perturbation analysis in Stochastic Hybrid Systems (SHS) of arbitrary structure. In particular, Infinitesimal Perturbation Analysis (IPA) is used to provide unbiased gradient estimates of performance metrics with respect to various controllable parameters. These can be combined with standard gradient-based algorithms for optimization purposes and ...

متن کامل

A Bi-objective Stochastic Optimization Model for Humanitarian Relief Chain by Using Evolutionary Algorithms

Due to the increasing amount of natural disasters such as earthquakes and floods and unnatural disasters such as war and terrorist attacks, Humanitarian Relief Chain (HRC) is taken into consideration of most countries. Besides, this paper aims to contribute humanitarian relief chains under uncertainty. In this paper, we address a humanitarian logistics network design problem including local dis...

متن کامل

Gradient Estimation for Stochastic Optimization of Optical Code-Division Multiple-Access Systems: Part II - Adaptive Detection

In this sequel, we develop infinitesimal perturbation analysis (IPA)-based stochastic gradient algorithms for deriving optimum detectors with the average probability of bit error being the objective function that is minimized. Specifically, we develop both a class of linear as well as nonlinear (threshold) detectors. In the linear scheme, the receiver despreads the received optical signal with ...

متن کامل

ذخیره در منابع من


  با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

عنوان ژورنال:
  • Discrete Event Dynamic Systems

دوره 10  شماره 

صفحات  -

تاریخ انتشار 2000