Competing particle systems evolving by interacting Lévy processes

DOI10.1214/10-AAP743MaRDI QIDQ655586zbMATH OpenFDO

Publication date 4 January 2012

Published in The Annals of Applied Probability (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/1002.2811

capital distributions Harris recurrence Lévy processes Lévy queueing networks semimartingale reflected Brownian motions stochastic differential equations

Mathematics Subject Classification ID

Processes with independent increments; Lévy processes (60G51) Stochastic ordinary differential equations (aspects of stochastic analysis) (60H10) Continuous-time Markov processes on general state spaces (60J25) Interacting random processes; statistical mechanics type models; percolation theory (60K35) Auctions, bargaining, bidding and selling, and other market models (91B26)

Abstract: We consider finite and infinite systems of particles on the real line and half-line evolving in continuous time. Hereby, the particles are driven by i.i.d. L'{e}vy processes endowed with rank-dependent drift and diffusion coefficients. In the finite systems we show that the processes of gaps in the respective particle configurations possess unique invariant distributions and prove the convergence of the gap processes to the latter in the total variation distance, assuming a bound on the jumps of the L'{e}vy processes. In the infinite case we show that the gap process of the particle system on the half-line is tight for appropriate initial conditions and same drift and diffusion coefficients for all particles. Applications of such processes include the modeling of capital distributions among the ranked participants in a financial market, the stability of certain stochastic queueing and storage networks and the study of the Sherrington--Kirkpatrick model of spin glasses.

Review text

Consider a system of competing particles indexed by \(i \in I\), where either \(I = \{1,\dots,N\}\) or \(I= \mathbb N\). The positions of the particles evolve on the real line \(\mathbb R\) or on the half line \([b,\infty)\) with barrier \(b \in \mathbb R\). These are called unregulated and regulated evolutions, respectively. Each particle \(i \in I\) moves according to the stochastic differential equation \[ d X^i_t = \sum_{j\in I} {\mathbf 1}_{\{ X^i_t = X^{(j)}_t\}}\delta_j dt + \sum_{j\in I} {\mathbf 1}_{\{ X^i_t = X^{(j)}_t\}}\sigma_j d W^i_t + dL^i_t + dR^i_t, \] where \(X^{(1)}_t \leq X^{(2)}_t \leq \ldots\) are the order statistics of \(X^1_t,X^2_t,\dots\), the driving processes \(W^i\), \(i \in I\), are independent standard Brownian motions, \(L^i\), \(i \in I\), are i.i.d. finite-activity pure-jump Lévy processes, and \(R^i\), \(i \in I\), are the processes that ensure that the particles do not cross the barrier \(b\) in the regulated case (in the unregulated case \(R^i\), \(i \in I\), are all equal to zero). The drifts \(\delta_i\), \(i \in I\), and the volatilities \(\sigma_i\), \(i\in I\), are constants. The object of interest in the paper is the process of gaps, defined by \[ Z = (X^{(2)}-X^{(1)},X^{(3)}-X^{(1)},\dots) \] for the unregulated evolution and by \[ Z = (X^{(1)}-b,X^{(2)}-b,\dots) \] for the regulated one. Under certain assumptions on the drifts, volatilities, and the driving Lévy processes, it is shown that when \(I\) is finite, the process \(Z\) has a unique invariant law and the law of \(Z_t\) converges to it in total variation as \(t \rightarrow \infty\) in both the unregulated and the regulated case. Additionally, when \(I = \mathbb N\), it is shown that the family \(\{Z_t : t \geq 0\}\) is tight in \(\mathbb R_+^{\mathbb N}\) for the regulated evolution.

Recommendations

Cites work

Cited in

(31)

This page was built for publication: Competing particle systems evolving by interacting Lévy processes

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q655586)