Scholarlink

Abstract <p>This work introduces a characteristic function-based training scheme for statistics-informed neural networks (SINNs) to model trawl processes–a class of ambit processes defined via Levy bases that capture complex temporal dependencies. The proposed approach learns finite-dimensional distributions directly without requiring external simulations, overcoming computational limitations of traditional methods, especially when closed-form expressions are unavailable. Numerical experiments, including applications to Ornstein–Uhlenbeck and gamma-trawl processes, demonstrate the method’s effectiveness in qualitative inference and its potential to accelerate Monte Carlo simulations for stochastic modeling tasks.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Simulation of Trawl Processes Using SINN Architectures

  • 期刊论文

  • 发表日期：

    2025年12月1日

查看全文
  • K. E. Belkova,
  • M. D. Mikhailov

摘要

Abstract

This work introduces a characteristic function-based training scheme for statistics-informed neural networks (SINNs) to model trawl processes–a class of ambit processes defined via Levy bases that capture complex temporal dependencies. The proposed approach learns finite-dimensional distributions directly without requiring external simulations, overcoming computational limitations of traditional methods, especially when closed-form expressions are unavailable. Numerical experiments, including applications to Ornstein–Uhlenbeck and gamma-trawl processes, demonstrate the method’s effectiveness in qualitative inference and its potential to accelerate Monte Carlo simulations for stochastic modeling tasks.