其他名称:
24-11-00123
项目负责人:
Pirogov Sergey
发表日期:
2024
主持机构:
National Research University Higher School of Economics,
国家:
俄罗斯
开始日期:
2024
结束日期:
2026
简介:
The project consists of three main areas.$Direction No. 1 - Stochastic dynamics of populations.$Homogeneous contact models of spatial dynamics of populations do not adequately reflect the actual spatial heterogeneity of biological and social processes. Therefore, there is a natural need to study contact processes in inhomogeneous or random media and contact processes on more general spaces than Euclidean, including graphs and manifolds. In addition to the applications of population dynamics to biological and sociological problems, we note the application to the dynamo problem, in particular, to the generation of strong magnetic fields in the photosphere of hot stars, and to the problem of modeling the temperature field in a turbulent current in the ocean (description of turbulent vortices leading to turbulent diffusion).$Direction No. 2 - Nonclassical parametrix of nonlinear diffusions and Markov chains and their applications.$We plan to consider non-classical parametrix methods and their applications. Namely, the asymptotic parametrix method will be developed when the diffusion matrices and trends of the pre-limit objects will only asymptotically coincide with the diffusion matrices and trends of the limit objects. This development of the method, in addition to theoretical interest, has practical applications, for example, in problems of mathematical statistics when constructing diffusion approximations for sums of induced ordered statistics. The parametrix operator method will be developed, which does not assume the existence of transient densities, but operates only with transient probabilities. Degenerate processes that do not have a transition density naturally arise, for example, as limiting ones in the multi-armed bandit problem, for such processes there is also the task of analyzing the weak error of discrete approximations of solutions. The classical parametrix method (i.e. assuming the existence of transition densities) "does not work" here. We also propose to develop a "non-Euclidean" parametrix method in hyperbolic spaces of arbitrary dimension. This will be a far - reaching generalisation of the classical work of F. Karpelevich, V. Tutubalin and M. Shur on random inhomogeneities in radio waveguides. , As the first step, the simple Markov chain will be considered, namely: "sums" in the sense of Mobius addition of independent random variables with values in the Poincare ball. For this model, the CLT, LLT, LLN and SLLN will be obtained.$Direction No. 3 - Analysis and limit properties of singular stochastic multi-particle systems and associated nonlinear diffusions of the McKean–Vlasov type.$The qualitative and quantitative study of large-population of stochastic particle systems with mean-field interaction is one of the major trend in modern stochastic modelling and stochastic analysis. These systems are closely connected to the class of McKean-Vlasov SDEs and heads to the rigorous design of numerical particle methods. These methods are used in a wide range of engineering problems from physics, biology, social, financial, economic and computer sciences. For this particular research line, we aim to address the two-fold problem of the wellposedness and the infinitely large population asymptotic of singular stochastic interacting particle systems. These questions are more specifically addressed in the general framework of distributional Besov-type mean-field interaction kernel and general stable kinetic noise. While practically relevant instances carry such irregularities - rendering complex aggregation mechanisms or turbulence phenemena -, these irregularities are often studied at the level of the McKean-Vlasov SDEs leaving the design of sharp particle approximation to be solved by a generic smoothening of the interactions. This smoothening is nevertheless known to severly alter intrinsic properties of the particle systems and large population asymptotic. It is expected that the results of the research line will, beyond their mathematical interest, enable to extend the current state of numerical particle methods.
专业领域:
其他
语种:
英语