Scholarlink

<p>This paper devises a novel neural-inspired approach for simulating multivariate extremes. Specifically, we propose a GAN-based generative model for sampling multivariate data exceeding large thresholds, giving rise to what we refer to as the ExceedGAN&#xa0;algorithm. Our approach is based on approximating marginal log-quantile functions using feedforward neural networks with eLU activation functions specifically introduced for bias correction. An error bound is provided on the margins, assuming a <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(\varvec{J}\)</EquationSource> </InlineEquation>th order condition from extreme value theory. The numerical experiments illustrate that ExceedGAN&#xa0;outperforms competitors, both on synthetic and real-world data sets.</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

ExceedGAN: simulation above extreme thresholds using Generative Adversarial Networks

  • 期刊论文

  • 发表日期：

    2026年1月28日

查看全文
  • Michaël Allouche,
  • Stéphane Girard,
  • Emmanuel Gobet

摘要

This paper devises a novel neural-inspired approach for simulating multivariate extremes. Specifically, we propose a GAN-based generative model for sampling multivariate data exceeding large thresholds, giving rise to what we refer to as the ExceedGAN algorithm. Our approach is based on approximating marginal log-quantile functions using feedforward neural networks with eLU activation functions specifically introduced for bias correction. An error bound is provided on the margins, assuming a \(\varvec{J}\) th order condition from extreme value theory. The numerical experiments illustrate that ExceedGAN outperforms competitors, both on synthetic and real-world data sets.