Enhancing neural operator learning with invariants to simultaneously learn various physical mechanisms  

作　　者：Siran Li Chong Liu Hao Ni 

机构地区：[1]School of Mathematical Sciences,Shanghai Jiao Tong University,China [2]Institute of Mathematical Sciences,ShanghaiTech University,China [3]Department of Mathematics,University College London,UK

出　　处：《National Science Review》2024年第8期42-43,共2页国家科学评论（英文版）

摘　　要：Partial differential equations(PDEs)play a fundamental role in the modelling and analysis of a wide range of physical and geometric problems.Numerous numerical techniques,classical and new,have been proposed to approximate PDE solutions,aiming at attaining high accuracy and efficiency.Most recently,by utilising deep neural networks to represent PDE solutions,machine learning(ML)methods have emerged as a revolutionary tool that demonstrates enormous potential to overcome the curse of dimensionality and to deal with complex geometries.

关 键 词：INVARIANTS OPERATOR OVERCOME 

分 类 号：TP39[自动化与计算机技术—计算机应用技术]