混合代理模型:有限冲击型不连续偏微分方程的 Gibbs 现象
图像与视频处理
2025-01-22 v2 计算机视觉与模式识别
机器学习
神经元与认知
摘要
我们引入混合代理模型 (HSMs) 的概念——将多变量多项式与 Heaviside 函数组合——作为具有有限多个跳跃不连续性函数的近似。我们利用 HSMs formulating 用于求解非正则偏微分方程 (PDEs) 的变分优化方法,后者具有非连续的冲击型解。HSM 技术同时获得冲击位置和高度的参数化以及 PDE 的解。我们表明 HSM 技术规克了经典数值方法所能达到的著名 Gibbs 现象限制。数值实验涵盖线性和非线性冲击传播,展示了 HSM 技术的强大逼近能力。
引用
@article{arxiv.2408.02496,
title = {Automatic rating of incomplete hippocampal inversions evaluated across multiple cohorts},
author = {Lisa Hemforth and Baptiste Couvy-Duchesne and Kevin De Matos and Camille Brianceau and Matthieu Joulot and Tobias Banaschewski and Arun L. W. Bokde and Sylvane Desrivières and Herta Flor and Antoine Grigis and Hugh Garavan and Penny Gowland and Andreas Heinz and Rüdiger Brühl and Jean-Luc Martinot and Marie-Laure Paillère Martinot and Eric Artiges and Dimitri Papadopoulos and Herve Lemaitre and Tomas Paus and Luise Poustka and Sarah Hohmann and Nathalie Holz and Juliane H. Fröhner and Michael N. Smolka and Nilakshi Vaidya and Henrik Walter and Robert Whelan and Gunter Schumann and Christian Büchel and JB Poline and Bernd Itterman and Vincent Frouin and Alexandre Martin and IMAGEN study group and Claire Cury and Olivier Colliot},
journal= {arXiv preprint arXiv:2408.02496},
year = {2025}
}
备注
Accepted for publication at the Journal of Machine Learning for Biomedical Imaging (MELBA) https://melba-journal.org/2024:016