High Order Deep Neural Network for Solving High Frequency Partial Differential Equations
作者机构:Qian Xuesen Laboratory of Space TechnologyChina Academy of Space TechnologyBeijing 100875P.R.China School of Mathematics and StatisticsWuhan UniversityWuhan 430072P.R.China Information Engineering UniversityZhengzhou 450001P.R.China
出 版 物:《Communications in Computational Physics》 (计算物理通讯(英文))
年 卷 期:2022年第31卷第2期
页 面:370-397页
核心收录:
学科分类:07[理学] 0701[理学-数学] 0702[理学-物理学] 070101[理学-基础数学]
基 金:supported partly by National Key R&D Program of China with grants 2019YFA0709600,2019YFA0709602 National Natural Science Foundation of China with grants 11831016,12101609 the Innovation Foundation of Qian Xuesen Laboratory of Space Technology
主 题:Deep neural network high order methods high frequency PDEs
摘 要:This paper proposes a high order deep neural network(HOrderDNN)for solving high frequency partial differential equations(PDEs),which incorporates the idea of“high orderfrom finite element methods(FEMs)into commonly-used deep neural networks(DNNs)to obtain greater approximation *** main idea of HOrderDNN is introducing a nonlinear transformation layer between the input layer and the first hidden layer to form a high order polynomial space with the degree not exceeding p,followed by a normal *** order p can be guided by the regularity of solutions of *** performance of HOrderDNNis evaluated on high frequency function fitting problems and high frequency Poisson and Helmholtz *** results demonstrate that:HOrderDNNs(p1)can efficiently capture the high frequency information in target functions;and when compared to physics-informed neural network(PINN),HOrderDNNs(p1)converge faster and achieve much smaller relative errors with same number of trainable *** particular,when solving the high frequency Helmholtz equation in Section 3.5,the relative error of PINN stays around 1 with its depth and width increase,while the relative error can be reduced to around 0.02 as p increases(see Table 5).