arXiv subjects

Peter W. Glynn

Publications and source records attributed to Peter W. Glynn.

53 records · Page 3Linked to original sources

arXiv2018

Affine Jump-Diffusions: Stochastic Stability and Limit Theorems

Affine jump-diffusions constitute a large class of continuous-time stochastic models that are particularly popular in finance and economics due to their analytical tractability. Methods for parameter estimation for such processes require ergodicity in order establish consistency and asymptotic normality of the associated estimators. In this paper, we develop stochastic stability conditions for affine jump-diffusions, thereby providing the needed large-sample theoretical support for estimating such processes. We establish ergodicity for such models by imposing a `strong mean reversion' condition and a mild condition on the distribution of the jumps, i.e. the finiteness of a logarithmic moment. Exponential ergodicity holds if the jumps have a finite moment of a positive order. In addition, we prove strong laws of large numbers and functional central limit theorems for additive functionals for this class of models.

q-fin.MF↗

arXiv2018

A $c/μ$-Rule for Service Resource Allocation in Group-Server Queues

In this paper, we study a dynamic on/off server scheduling problem in a queueing system with multi-class servers, where servers are heterogeneous and can be classified into $K$ groups. Servers in the same group are homogeneous. A scheduling policy determines the number of working servers (servers that are turned on) in each group at every state $n$ (number of customers in the system). Our goal is to find the optimal scheduling policy to minimize the long-run average cost, which consists of an increasing convex holding cost and a linear operating cost. We use the sensitivity-based optimization theory to characterize the optimal policy. A necessary and sufficient condition of the optimal policy is derived. We also prove that the optimal policy has monotone structures and a quasi bang-bang control is optimal. We find that the optimal policy is indexed by the value of $c - μG(n)$, where $c$ is the operating cost rate, $μ$ is the service rate for a server, and $G(n)$ is a computable quantity called perturbation realization factor. Specifically, the group with smaller negative $c - μG(n)$ is more preferred to be turned on, while the group with positive $c - μG(n)$ should be turned off. However, the preference ranking of each group is affected by $G(n)$ and the preference order may change with the state $n$, the arrival rate, and the cost function. Under a reasonable condition of scale economies, we further prove that the optimal policy obeys a so-called $c$/$μ$-rule. That is, the servers with smaller $c$/$μ$ should be turned on with higher priority and the preference order of groups remains unchanged. This rule can be viewed as a sister version of the famous $cμ$-rule for polling queues. With the monotone property of $G(n)$, we further prove that the optimal policy has a multi-threshold structure when the $c$/$μ$-rule is applied.

math.OC↗

arXiv2018

A Probabilistic Proof of the Perron-Frobenius Theorem

The Perron-Frobenius theorem plays an important role in many areas of management science and operations research. This paper provides a probabilistic perspective on the theorem, by discussing a proof that exploits a probabilistic representation of the Perron-Frobenius eigenvalue and eigenvectors in terms of the dynamics of a Markov chain. The proof provides conditions in both the finite-dimensional and infinite-dimensional settings under which the Perron-Frobenius eigenvalue and eigenvectors exist. Furthermore, the probabilistic representations that arise can be used to produce a Monte Carlo algorithm for computing the Perron-Frobenius eigenvalue and eigenvectors that will be explored elsewhere.

math.PR↗

arXiv2018

Approximating Systems Fed by Poisson Processes with Rapidly Changing Arrival Rates

This paper introduces a new asymptotic regime for simplifying stochastic models having non-stationary effects, such as those that arise in the presence of time-of-day effects. This regime describes an operating environment within which the arrival process to a service system has an arrival intensity that is fluctuating rapidly. We show that such a service system is well approximated by the corresponding model in which the arrival process is Poisson with a constant arrival rate. In addition to the basic weak convergence theorem, we also establish a first order correction for the distribution of the cumulative number of arrivals over $[0,t]$, as well as the number-in-system process for an infinite-server queue fed by an arrival process having a rapidly changing arrival rate. This new asymptotic regime provides a second regime within which non-stationary stochastic models can be reasonably approximated by a process with stationary dynamics, thereby complementing the previously studied setting within which rates vary slowly in time.

math.PR↗

arXiv2018

Approximating Performance Measures for Slowly Changing Non-stationary Markov Chains

This paper is concerned with the development of rigorous approximations to various expectations associated with Markov chains and processes having non-stationary transition probabilities. Such non-stationary models arise naturally in contexts in which time-of-day effects or seasonality effects need to be incorporated. Our approximations are valid asymptotically in regimes in which the transition probabilities change slowly over time. Specifically, we develop approximations for the expected infinite horizon discounted reward, the expected reward to the hitting time of a set, the expected reward associated with the state occupied by the chain at time $n$, and the expected cumulative reward over an interval $[0,n]$. In each case, the approximation involves a linear system of equations identical in form to that which one would need to solve to compute the corresponding quantity for a Markov model having stationary transition probabilities. In that sense, the theory provides an approximation no harder to compute than in the traditional stationary context. While most of the theory is developed for finite state Markov chains, we also provide generalizations to continuous state Markov chains, and finite state Markov jump processes in continuous time. In the latter context, one of our approximations coincides with the uniform acceleration asymptotic due to Massey and Whitt (1998).

math.PR↗

arXiv2018

Likelihood Ratio Gradient Estimation for Steady-State Parameters

We consider a discrete-time Markov chain $\boldsymbolΦ$ on a general state-space ${\sf X}$, whose transition probabilities are parameterized by a real-valued vector $\boldsymbolθ$. Under the assumption that $\boldsymbolΦ$ is geometrically ergodic with corresponding stationary distribution $π(\boldsymbolθ)$, we are interested in estimating the gradient $\nabla α(\boldsymbolθ)$ of the steady-state expectation $$α(\boldsymbolθ) = π( \boldsymbolθ) f.$$ To this end, we first give sufficient conditions for the differentiability of $α(\boldsymbolθ)$ and for the calculation of its gradient via a sequence of finite horizon expectations. We then propose two different likelihood ratio estimators and analyze their limiting behavior.

math.PR↗

arXiv2017

On Gaussian Limits and Large Deviations for Queues Fed by High Intensity Randomly Scattered Traffic

We study a single server FIFO queue that offers general service. Each of n customers enter the queue at random time epochs that are inde- pendent and identically distributed. We call this the random scattering traffic model, and the queueing model RS/G/1. We study the workload process associated with the queue in two different settings. First, we present Gaussian process approximations in a high intensity asymptotic scale and characterize the transient distribution of the approximation. Second, we study the rare event paths of the workload by proving a large deviations principle in the same high intensity regime. We also obtain exact asymptotics for the Gaussian approximations developed prior. This analysis significantly extends and simplifies recent work in [1] on uniform population acceleration asymptotics to the queue length and workload in the RS/G/1 queue.

math.PR↗

arXiv2016

A Generalized Fundamental Matrix for Computing Fundamental Quantities of Markov Systems

As is well known, the fundamental matrix $(I - P + e π)^{-1}$ plays an important role in the performance analysis of Markov systems, where $P$ is the transition probability matrix, $e$ is the column vector of ones, and $π$ is the row vector of the steady state distribution. It is used to compute the performance potential (relative value function) of Markov decision processes under the average criterion, such as $g=(I - P + e π)^{-1} f$ where $g$ is the column vector of performance potentials and $f$ is the column vector of reward functions. However, we need to pre-compute $π$ before we can compute $(I - P + e π)^{-1}$. In this paper, we derive a generalization version of the fundamental matrix as $(I - P + e r)^{-1}$, where $r$ can be any given row vector satisfying $r e \neq 0$. With this generalized fundamental matrix, we can compute $g=(I - P + e r)^{-1} f$. The steady state distribution is computed as $π= r(I - P + e r)^{-1}$. The Q-factors at every state-action pair can also be computed in a similar way. These formulas may give some insights on further understanding how to efficiently compute or estimate the values of $g$, $π$, and Q-factors in Markov systems, which are fundamental quantities for the performance optimization of Markov systems.

math.OC↗

arXiv2015

Measuring the Initial Transient: Reflected Brownian Motion

We analyze the convergence to equilibrium of one-dimensional reflected Brownian motion (RBM) and compute a number of related initial transient formulae. These formulae are of interest as approximations to the initial transient for queueing systems in heavy traffic, and help us to identify settings in which initialization bias is significant. We conclude with a discussion of mean square error for RBM. Our analysis supports the view that initial transient effects for RBM and related models are typically of modest size relative to the intrinsic stochastic variability, unless one chooses an especially poor initialization.

stat.ME↗

arXiv2014

Exact Estimation for Markov Chain Equilibrium Expectations

We introduce a new class of Monte Carlo methods, which we call exact estimation algorithms. Such algorithms provide unbiased estimators for equilibrium expectations associated with real- valued functionals defined on a Markov chain. We provide easily implemented algorithms for the class of positive Harris recurrent Markov chains, and for chains that are contracting on average. We further argue that exact estimation in the Markov chain setting provides a significant theoretical relaxation relative to exact simulation methods.

stat.CO↗

arXiv2014

Central Limit Theorems and Large Deviations for Additive Functionals of Reflecting Diffusion Processes

This paper develops central limit theorems (CLT's) and large deviations results for additive functionals associated with reflecting diffusions in which the functional may include a term associated with the cumulative amount of boundary reflection that has occurred. Extending the known central limit and large deviations theory for Markov processes to include additive functionals that incorporate boundary reflection is important in many applications settings in which reflecting diffusions arise, including queueing theory and economics. In particular, the paper establishes the partial differential equations that must be solved in order to explicitly compute the mean and variance for the CLT, as well as the associated rate function for the large deviations principle.

math.PR↗

arXiv2013

Shape-constrained Estimation of Value Functions

We present a fully nonparametric method to estimate the value function, via simulation, in the context of expected infinite-horizon discounted rewards for Markov chains. Estimating such value functions plays an important role in approximate dynamic programming and applied probability in general. We incorporate "soft information" into the estimation algorithm, such as knowledge of convexity, monotonicity, or Lipchitz constants. In the presence of such information, a nonparametric estimator for the value function can be computed that is provably consistent as the simulated time horizon tends to infinity. As an application, we implement our method on price tolling agreement contracts in energy markets.

math.PR↗

arXiv2013

Exact Simulation of Non-stationary Reflected Brownian Motion

This paper develops the first method for the exact simulation of reflected Brownian motion (RBM) with non-stationary drift and infinitesimal variance. The running time of generating exact samples of non-stationary RBM at any time $t$ is uniformly bounded by $\mathcal{O}(1/\barγ^2)$ where $\barγ$ is the average drift of the process. The method can be used as a guide for planning simulations of complex queueing systems with non-stationary arrival rates and/or service time.

math.PR↗

arXiv2012

A new approach to unbiased estimation for SDE's

In this paper, we introduce a new approach to constructing unbiased estimators when computing expectations of path functionals associated with stochastic differential equations (SDEs). Our randomization idea is closely related to multi-level Monte Carlo and provides a simple mechanism for constructing a finite variance unbiased estimator with "square root convergence rate" whenever one has available a scheme that produces strong error of order greater than 1/2 for the path functional under consideration.

q-fin.CP↗

arXiv2011

On the Convergence of Finite Order Approximations of Stationary Time Series

The approximation of a stationary time-series by finite order autoregressive (AR) and moving averages (MA) is a problem that occurs in many applications. In this paper we study asymptotic behavior of the spectral density of finite order approximations of wide sense stationary time series. It is shown that when the on the spectral density is non-vanishing in $[-π,π]$ and the covariance is summable, the spectral density of the approximating autoregressive sequence converges at the origin. Under additional mild conditions on the coefficients of the Wold decomposition it is also shown that the spectral densities of both moving average and autoregressive approximations converge in $L_2$ as the order of approximation increases.

math.ST↗

arXiv2011

Uniform Approximations for the M/G/1 Queue with Subexponential Processing Times

This paper studies the asymptotic behavior of the steady-state waiting time, W_infty, of the M/G/1 queue with subexponenential processing times for different combinations of traffic intensities and overflow levels. In particular, we provide insights into the regions of large deviations where the so-called heavy traffic approximation and heavy tail asymptotic hold. For queues whose service time distribution decays slower than e^{-sqrt{t}} we identify a third region of asymptotics where neither the heavy traffic nor the heavy tailed approximations are valid. These results are obtained by deriving approximations for P(W_infty > x) that are either uniform in the traffic intensity as the tail value goes to infinity or uniform on the positive axis as the traffic intensity converges to one. Our approach makes clear the connection between the asymptotic behavior of the steady-state waiting time distribution and that of an associated random walk.

math.PR↗

arXiv2006

Tail asymptotics for the maximum of perturbed random walk

Consider a random walk $S=(S_n:n\geq 0)$ that is ``perturbed'' by a stationary sequence $(ξ_n:n\geq 0)$ to produce the process $(S_n+ξ_n:n\geq0)$. This paper is concerned with computing the distribution of the all-time maximum $M_{\infty}=\max \{S_k+ξ_k:k\geq0\}$ of perturbed random walk with a negative drift. Such a maximum arises in several different applications settings, including production systems, communications networks and insurance risk. Our main results describe asymptotics for $\mathbb{P}(M_{\infty}>x)$ as $x\to\infty$. The tail asymptotics depend greatly on whether the $ξ_n$'s are light-tailed or heavy-tailed. In the light-tailed setting, the tail asymptotic is closely related to the Cramér--Lundberg asymptotic for standard random walk.

math.PR↗