Levy processes and their applications

征收流程及其应用

• 批准号: RGPIN-2019-06320
• 负责人: Kuznetsov, Alexey
• 金额: $ 1.82万
• 依托单位: York University
• 依托单位国家: 加拿大
• 项目类别: Discovery Grants Program - Individual
• 财政年份: 2019
• 资助国家: 加拿大
• 起止时间: 2019-01-01 至 2020-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=680846
• 关键词: Levy; processes; their; applications

Imagine a particle that travels along the line in the following way: At each moment of time the particle decides randomly (and independently of the past) whether to jump to the left or to the right. Mathematicians would call this simple model a "discrete time random walk". A natural generalisation of this model to continuous time would be called a "one-dimensional Levy process". The rich class Levy of processes occupies the central stage in much of the theory of stochastic processes. Levy processes are indispensable in the study of fine properties of many important objects in pure probability, such as branching processes, random trees, fragmentation processes and self-similar Markov processes. They are also all-important in many applied probability models, in particular in such areas as queuing theory and optimal control, mathematical finance and actuarial mathematics.***Levy processes have been studied from 1940s, with periods of heightened interest in 1960s- early 1970s and after late 1990s. However, there are still many important unresolved problems in this area. One of these problems concerns investigating how a stable process process exits from an interval (a stable process is the only Levy process that is self-similar: its law is preserved under a simultaneous scaling of time and space). Making any progress in this area would be an important achievement and would lead to advances in many areas where stable processes are applied. I intend to use recent results on Wiener-Hopf factorization for matrices to study this problem using a mix of complex-analytical and probabilistic methods. Another area of intense current activity is the study of Generalized Gamma Convolutions (GGC) -- a very useful class of distributions that has a lot of analytical structure and that includes many distributions used in applications (lognormal, Weibull, Pareto, etc.). Here I plan to focus on developing numerical methods for working with this class of distributions, in particular, methods for computing and approximating Laplace transforms of these random variables. I also plan to investigate multi-dimensional generalisations of the GGC class and to study dependence structures and copulas that arise in this way and apply them in Actuarial Science and Mathematical Finance. The third direction of my future research will be about approximating arbitrary Levy processes by simpler, but more computationally efficient processes. I will consider processes with jumps of rational transform (these are the easiest processes for computational purposes) and will develop algorithms for approximating any Levy process by these ones. I believe that this work will be useful for all practitioners and applied mathematicians who are using Levy processes for modelling purposes. **

想象一个粒子以以下方式沿着这条线运动：在每一个时刻，粒子随机决定（并且独立于过去）是向左还是向右跳。数学家将这种简单的模型称为“离散时间随机漫步”。将这个模型自然推广到连续时间将被称为“一维列维过程”。在随机过程的许多理论中，富有阶层的列维占有中心地位。在纯概率中研究分支过程、随机树、碎片过程和自相似马尔可夫过程等许多重要对象的精细性质时，Levy过程是必不可少的。它们在许多应用概率模型中也非常重要，特别是在排队论和最优控制、数学金融和精算数学等领域。***利维过程从20世纪40年代开始研究，在20世纪60年代至70年代初和90年代末之后的时期引起了极大的兴趣。然而，在这一领域仍有许多重要问题尚未解决。其中一个问题是研究一个稳定过程如何从一个区间中退出（稳定过程是唯一自相似的Levy过程：它的定律在时间和空间的同时缩放下保持不变）。在这一领域取得任何进展都将是一项重要的成就，并将导致在应用稳定过程的许多领域取得进展。我打算利用最近关于矩阵的维纳-霍普夫分解的结果，使用复杂分析和概率方法的混合来研究这个问题。另一个热点领域是对广义伽马卷积（GGC）的研究——这是一类非常有用的分布，它有很多分析结构，包括许多应用中使用的分布（对数正态分布、威布尔分布、帕累托分布等）。在这里，我计划重点发展处理这类分布的数值方法，特别是计算和近似这些随机变量的拉普拉斯变换的方法。我还计划研究GGC类的多维推广，并研究由此产生的依赖结构和copula，并将其应用于精算科学和数学金融。我未来研究的第三个方向将是通过更简单，但计算效率更高的过程来近似任意Levy过程。我将考虑具有有理变换跳跃的过程（这些是用于计算目的的最简单的过程），并将开发用这些过程近似任何列维过程的算法。我相信这项工作将对所有使用Levy过程进行建模的实践者和应用数学家有用。**