Matrix Exponential Distributions Research Articles

New LST of inter-departure times in PH/G/1 queue, and extensions to ME/G/1 and G/G/1 queues

In this paper, we provide a new approach to model the inter-departure times distribution in a PH/G/1 queue. This approach enables to further model the inter-departure times distribution in more general queues as well. Initially, we propose to express the Laplace–Stieltjes transform (LST) of inter-departure times in PH/G/1 queues by exploiting the probabilistic interpretation of phase-type distributions. Using this interpretation enables to eliminate the necessity of the matrix-geometric method, and thus significantly reduces the computational complexity. Then, we use the LST of inter-departure times distribution in a Cm/G/1 queue to express this LST in a ME/G/1 queue, where ME is a Matrix-Exponential distribution. We validate it in a few ME/G/1 examples. Finally, we propose to approximate the LST of inter-departure times distribution in a G/G/1 queue by employing the above LST of the proper PH/G/1 queue. Without loss of generality, we demonstrate our proposed approximation by using the LST as obtained in a Cm/G/1 queue, while illustrating by a few G/G/1 examples that the accuracy can be as good as one might want.

An Optimal Inverse Laplace Transform Method Without Positive and Negative Overshoot – An Integral Based Interpretation

We propose a numerical inverse Laplace transformation method without overshoot which is derived from matrix exponential (ME) distributions with minimal coefficient of variation. We discuss the properties of the method through an integral based interpretation of numerical inverse Laplace transformation methods belonging to the Abate–Whitt framework. Compared to the previously applied non-overshooting alternative, the “Erlang” method, the error of the proposed method improves from O(1/n) to O(1/n2) while it maintains the same computational complexity.

On the Resolvent of the Lévy Process with Matrix-Exponential Distribution of Jumps

We consider the representations of resolvent for a Levy process whose jumps have a matrix-exponential distribution.

First exit from an open set for a matrix-exponential Lévy process

We study the first exit from a general open set for a one-dimensional Lévy process, where the Lévy measure is proportional to a two-sided matrix-exponential distribution. Under appropriate conditions on the Lévy measure, we obtain an explicit solution for the joint distribution of the first-exit time and the position of the Lévy process upon first exit, in terms of the zeros and poles of the corresponding Laplace exponent. The present result complements several earlier works on the use of exit sets for Lévy processes with algebraically similar Laplace exponents, where exits from open intervals are the main focus.

Throughput Analysis of Hybrid-ARQ—A Matrix Exponential Distribution Approach

We propose a novel performance analysis framework for lossless- and truncated-hybrid automatic repeat request (HARQ) that enables neat, general, closed-form throughput expressions in a matrix exponential (ME) distribution form. This approach is applicable to all HARQ schemes for which the probability density function of the effective channel can be characterized by a rational Laplace transform, or equivalently, an ME-distribution. This includes, for example, repetition redundancy HARQ in ME distributed channels. Throughput expressions are also given for the $K$ -truncated-HARQ $N$ -fold diversity, ARQ $N$ -fold diversity, and lossless-HARQ 2-fold diversity cases in the ME distributed channel. Schemes with effective channels of non-rational Laplace transforms, such as IR-HARQ, are explored using truncated continued fractions. A novel integration trick is developed for the integration of ME distributions with singular matrices and yields the simple throughput expression of lossless-HARQ. We also give general analytical expressions for the optimal throughput and optimal rate point that benefit from the compact ME-distribution form proposed.

On the use of functional calculus for phase-type and related distributions

The area of phase-type distributions is renowned for its ability to obtain closed form formulas or algorithmically exact solutions to many complex stochastic models. The method of functional calculus will provide an additional tool along these lines for establishing results in terms of functions of matrices. Functional calculus, which is a branch of operator theory frequently associated with complex analysis, can be applied to phase-type and matrix-exponential distributions in a rather straightforward way. In this article we provide a number of examples of how to execute the formal arguments.

An alternative method for computing system-length distributions of BMAP/R/1 and BMAP/D/1 queues using roots

In this paper, we present closed-form expressions for system-length distributions in terms of roots outside the unit disk of the characteristic equation of the BMAP/R/1 queue, where arrival process is batch Markovian arrival process (BMAP) and R represents a class of distributions having rational Laplace–Stieltjes transform. The unknown boundary vector has been evaluated using the roots (inside and on the unit disk) of the characteristic equation. Several numerical results are presented for a variety of arrival and service-time distributions including phase-type (PH) and matrix-exponential (ME) which cover a wide variety of distributions that arise in applications. Using an approximated representation of ME distribution, results for BMAP/D/1 queue are also presented. We compared our result with the results obtained using classical matrix analytic method as well as cyclic reduction algorithm. It is shown that the computation-time of the proposed method does not depend upon the parameters like traffic intensity and correlation co-efficient. The method is analytically quite simple and easy to implement.

Phase Type Distributions with Finite Support

This research is motivated by the fact that many random variables of practical interest have a finite support. For fixed a < b, we consider the distribution of a random variable X = (a + Ymod(b − a)), where Y is a phase type (PH) random variable. We demonstrate that as we traverse for Y the entire set of PH distributions (or even any subset thereof like Coxian that is dense in the class of distributions on [0, ∞)), we obtain a class of matrix exponential distributions dense in (a, b). We call these Finite Support Phase Type Distributions (FSPH) of the first kind. A simple example shows that though dense, this class by itself is not very efficient for modeling; therefore, we introduce (and derive the EM algorithms for) two other classes of finite support phase type distributions (FSPH). The properties of denseness, connection to Markov chains, the EM algorithm, and ability to exploit matrix-based computations should all make these classes of distributions attractive not only for applied probability but also for a much wider variety of fields using statistical methodologies.

Queues and Risk Processes with Dependencies

We study the generalization of the G/G/1 queue obtained by relaxing the assumption of independence between inter-arrival times and service requirements. The analysis is carried out for the class of multivariate matrix exponential distributions introduced in Ref.[13]. In this setting, we obtain the steady-state waiting time distribution, and we show that the classical relation between the steady-state waiting time and workload distributions remains valid when the independence assumption is relaxed. We also prove duality results with the ruin functions in an ordinary and a delayed ruin process. These extend several known dualities between queueing and risk models in the independent case. Finally, we show that there exist stochastic order relations between the waiting times under various instances of correlation.

Constructing Matrix Exponential Distributions by Moments and Behavior around Zero

This paper deals with moment matching of matrix exponential (ME) distributions used to approximate general probability density functions (pdf). A simple and elegant approach to this problem is applying Padé approximation to the moment generating function of the ME distribution. This approach may, however, fail if the resulting ME function is not a proper probability density function; that is, it assumes negative values. As there is no known, numerically stable method to check the nonnegativity of general ME functions, the applicability of Padé approximation is limited to low-order ME distributions or special cases. In this paper, we show that the Padé approximation can be extended to capture the behavior of the original pdf around zero and this can help to avoid representations with negative values and to have a better approximation of the shape of the original pdf. We show that there exist cases when this extension leads to ME function whose nonnegativity can be verified, while the classical approach results in improper pdf. We apply the ME distributions resulting from the proposed approach in stochastic models and show that they can yield more accurate results.

Numerical analysis of rational processes beyond Markov chains

Matrix exponential distributions and rational arrival processes have been proposed as an extension to pure Markov models. The paper presents an approach where these process types are used to describe the timing behavior in quantitative models like queueing networks, stochastic Petri nets or stochastic automata networks. The resulting stochastic process, which is called a rational process, is defined and it is shown that the matrix governing the behavior of the process has a structured representation which allows one to represent the matrix in a very compact form.The main emphasis of the paper is on numerical techniques to compute the stationary distribution. It is shown which techniques from Markov chain analysis may also be applied for rational processes. Aggregation steps can be used to build an aggregated process which is a Markov chain and allows one to compute certain marginal probabilities for the detailed rational process. Furthermore, a block iteration method is presented which exploits the matrix structure and is an extension of previously published techniques for Markov chains.Finally, we present some examples showing that the extension of Markov models results in models which may have a significantly smaller state space than equivalently or similar behaving Markov chains.

Rational Automata Networks: A Non-Markovian Modeling Approach

A new class of non-Markovian models is introduced that results from the combination of stochastic automata networks and a very general class of stochastic processes, namely, rational arrival processes, which are derived from matrix exponential distributions. It is shown that the modeling formalism allows a compact representation of complex models with large state spaces. The resulting stochastic process is non-Markovian, but it can be analyzed with numerical techniques like a Markov chain, and the results at the level of the automata are stochastic distributions that can be used to compute standard performance and dependability results. The model class includes stochastic automata networks with phase-type distributed and correlated event times and also includes models that have a finite state space but cannot be represented by finite Markov chains. The paper introduces the model class, shows how the descriptor matrix can be represented in compact form, presents some example models, and outlines methods to analyze the new models.

A semi-infinite programming approach to identifying matrix-exponential distributions

The Laplace–Stieltjes transform of a matrix-exponential (ME) distribution is a rational function where at least one of its poles of maximal real part is real and negative. The coefficients of the numerator polynomial, however, are more difficult to characterise. It is known that they are contained in a bounded convex set that is the intersection of an uncountably infinite number of linear half-spaces. In order to determine whether a given vector of numerator coefficients is contained in this set (i.e. the vector corresponds to an ME distribution) we present a semi-infinite programming algorithm that minimises a convex distance function over the set. In addition, in the event that the given vector does not correspond to an ME distribution, the algorithm returns a closest vector which does correspond to one.

Characterizing matrix-exponential distributions of order 4 (abstract only)

The Laplace-Stieltjes transform of a matrix-exponential distribution is a rational function. If there are no common factors between the numerator and denominator polynomials, then the order of the matrix-exponential distribution is the degree of the denominator polynomial. Given a rational Laplace-Stieltjes transform, it is unknown, in general, when it corresponds to a matrix-exponential distribution. Matrix-exponential distributions of order 3 have been completely characterized in this manner, but in this talk we look at the problem of characterizing matrix-exponential distributions of order 4.

Acceptance-rejection methods for generating random variates from matrix exponential distributions and rational arrival processes (abstract only)

Stochastic models based on matrix exponential structures, like matrix exponential distributions and rational arrival processes, have gained popularity in analytical models recently. However the application of these models in simulation based evaluations is not as widespread yet. One of the possible reasons is the lack of efficient random variates generation methods. In this paper we propose methods for efficient random variates generation for matrix exponential stochastic models based on appropriate representations of the models.

Bilateral matrix-exponential distributions (abstract only)

In this article we define the classes of bilateral and multivariate bilateral matrix-exponential distributions. These distributions have support on the entire real space and have rational moment-generating functions. These distributions extend the class of bilateral phasetype distributions of [1] and the class of multivariate matrix-exponential distributions of [9]. We prove a characterization theorem stating that a random variable has a bilateral multivariate distribution if and only if all linear combinations of the coordinates have a univariate bilateral matrix-exponential distribution. As an application we demonstrate that certain multivariate disions, which are governed by the underlying Markov jump process generating a phasetype distribution, have a bilateral matrix-exponential distribution at the time of absorption, see also [4].

Rational Processes Related to Communicating Markov Processes

We define a class of stochastic processes, denoted as marked rational arrival processes (MRAPs), which is an extension of matrix exponential distributions and rational arrival processes. Continuous-time Markov processes with labeled transitions are a subclass of this more general model class. New equivalence relations between processes are defined, and it is shown that these equivalence relations are natural extensions of strong and weak lumpability and the corresponding bisimulation relations that have been defined for Markov processes. If a general rational process is equivalent to a Markov process, it can be used in numerical analysis techniques instead of the Markov process. This observation allows one to apply MRAPs like Markov processes and since the new equivalence relations are more general than lumpability and bisimulation, it is sometimes possible to find smaller representations of given processes. Finally, we show that the equivalence is preserved by the composition of MRAPs and can therefore be exploited in compositional modeling.

Rational Processes Related to Communicating Markov Processes

We define a class of stochastic processes, denoted as marked rational arrival processes (MRAPs), which is an extension of matrix exponential distributions and rational arrival processes. Continuous-time Markov processes with labeled transitions are a subclass of this more general model class. New equivalence relations between processes are defined, and it is shown that these equivalence relations are natural extensions of strong and weak lumpability and the corresponding bisimulation relations that have been defined for Markov processes. If a general rational process is equivalent to a Markov process, it can be used in numerical analysis techniques instead of the Markov process. This observation allows one to apply MRAPs like Markov processes and since the new equivalence relations are more general than lumpability and bisimulation, it is sometimes possible to find smaller representations of given processes. Finally, we show that the equivalence is preserved by the composition of MRAPs and can therefore be exploited in compositional modeling.

On the absolute ruin problem in a Sparre Andersen risk model with constant interest

In this paper, we extend the work of Mitric and Sendova (2010), which considered the absolute ruin problem in a risk model with debit and credit interest, to renewal and non-renewal structures. Our first results apply to MAP processes, which we later restrict to the Sparre Andersen renewal risk model with interclaim times that are generalized Erlang ( n) distributed and claim amounts following a Matrix-Exponential (ME) distribution (see for e.g. Asmussen and O’Cinneide (1997)). Under this scenario, we present a general methodology to analyze the Gerber–Shiu discounted penalty function defined at absolute ruin, as a solution of high-order linear differential equations with non-constant coefficients. Closed-form solutions for some absolute ruin related quantities in the generalized Erlang (2) case complement the results obtained under the classical risk model by Mitric and Sendova (2010).

Moment Distributions of Phase Type

Moment distributions of phase-type and matrix-exponential distributions are shown to remain within their respective classes. We provide a probabilistic phase-type representation for the former case and an alternative representation, with an analytically appealing form, for the latter. First order moment distributions are of special interest in areas like demography and economics, and we calculate explicit formulas for the Lorenz curve and Gini index used in these disciplines.