Pub Date : 2023-12-27DOI: 10.1007/s00211-023-01382-8
Carsten Carstensen, Sophie Puttkammer
Guaranteed lower Dirichlet eigenvalue bounds (GLB) can be computed for the m-th Laplace operator with a recently introduced extra-stabilized nonconforming Crouzeix–Raviart ((m=1)) or Morley ((m=2)) finite element eigensolver. Striking numerical evidence for the superiority of a new adaptive eigensolver motivates the convergence analysis in this paper with a proof of optimal convergence rates of the GLB towards a simple eigenvalue. The proof is based on (a generalization of) known abstract arguments entitled as the axioms of adaptivity. Beyond the known a priori convergence rates, a medius analysis is enfolded in this paper for the proof of best-approximation results. This and subordinated (L^2) error estimates for locally refined triangulations appear of independent interest. The analysis of optimal convergence rates of an adaptive mesh-refining algorithm is performed in 3D and highlights a new version of discrete reliability.
{"title":"Adaptive guaranteed lower eigenvalue bounds with optimal convergence rates","authors":"Carsten Carstensen, Sophie Puttkammer","doi":"10.1007/s00211-023-01382-8","DOIUrl":"https://doi.org/10.1007/s00211-023-01382-8","url":null,"abstract":"<p>Guaranteed lower Dirichlet eigenvalue bounds (GLB) can be computed for the <i>m</i>-th Laplace operator with a recently introduced extra-stabilized nonconforming Crouzeix–Raviart (<span>(m=1)</span>) or Morley (<span>(m=2)</span>) finite element eigensolver. Striking numerical evidence for the superiority of a new adaptive eigensolver motivates the convergence analysis in this paper with a proof of optimal convergence rates of the GLB towards a simple eigenvalue. The proof is based on (a generalization of) known abstract arguments entitled as the axioms of adaptivity. Beyond the known a priori convergence rates, a medius analysis is enfolded in this paper for the proof of best-approximation results. This and subordinated <span>(L^2)</span> error estimates for locally refined triangulations appear of independent interest. The analysis of optimal convergence rates of an adaptive mesh-refining algorithm is performed in 3D and highlights a new version of discrete reliability.</p>","PeriodicalId":49733,"journal":{"name":"Numerische Mathematik","volume":null,"pages":null},"PeriodicalIF":2.1,"publicationDate":"2023-12-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139053757","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"数学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-12-27DOI: 10.1007/s00211-023-01387-3
Gregor Gantner, Rob Stevenson
We consider the first-order system space–time formulation of the heat equation introduced by Bochev and Gunzburger (in: Bochev and Gunzburger (eds) Applied mathematical sciences, vol 166, Springer, New York, 2009), and analyzed by Führer and Karkulik (Comput Math Appl 92:27–36, 2021) and Gantner and Stevenson (ESAIM Math Model Numer Anal 55(1):283–299 2021), with solution components ((u_1,textbf{u}_2)=(u,-nabla _textbf{x} u)). The corresponding operator is boundedly invertible between a Hilbert space U and a Cartesian product of (L_2)-type spaces, which facilitates easy first-order system least-squares (FOSLS) discretizations. Besides (L_2)-norms of (nabla _textbf{x} u_1) and (textbf{u}_2), the (graph) norm of U contains the (L_2)-norm of (partial _t u_1 +{{,textrm{div},}}_textbf{x} textbf{u}_2). When applying standard finite elements w.r.t. simplicial partitions of the space–time cylinder, estimates of the approximation error w.r.t. the latter norm require higher-order smoothness of (textbf{u}_2). In experiments for both uniform and adaptively refined partitions, this manifested itself in disappointingly low convergence rates for non-smooth solutions u. In this paper, we construct finite element spaces w.r.t. prismatic partitions. They come with a quasi-interpolant that satisfies a near commuting diagram in the sense that, apart from some harmless term, the aforementioned error depends exclusively on the smoothness of (partial _t u_1 +{{,textrm{div},}}_textbf{x} textbf{u}_2), i.e., of the forcing term (f=(partial _t-Delta _x)u). Numerical results show significantly improved convergence rates.
我们考虑 Bochev 和 Gunzburger 引入的热方程一阶系统时空表述(见 Bochev 和 Gunzburger(编)《应用数学科学》第 166 卷,施普林格出版社,纽约,2009 年),以及 Führer 和 Karkulik 对其进行的分析(《计算数学应用》,纽约,2009 年):Führer and Karkulik (Comput Math Appl 92:27-36, 2021) and Gantner and Stevenson (ESAIM Math Model Numer Anal 55(1):283-299 2021) 对其进行了分析,其解分量为 ((u_1,textbf{u}_2)=(u,-nabla _textbf{x} u))。相应的算子在希尔伯特空间 U 和 (L_2)-type 空间的笛卡尔乘积之间是有界可逆的,这便于一阶系统最小二乘(FOSLS)离散化。除了(nabla _textbf{x} u_1) 和(textbf{u}_2)的(L_2)-规范外,U的(图)规范还包含(partial _t u_1 +{,textrm{div},}}_textbf{x} textbf{u}_2)的(L_2)-规范。当应用标准有限元对时空圆柱体进行简分时,对后一种规范的近似误差估计需要 (textbf{u}_2) 的高阶平稳性。在均匀分区和自适应细化分区的实验中,非光滑解 u 的收敛率低得令人失望。它们带有一个准内插值,该准内插值满足近似换向图的意义,即除了一些无害项之外,上述误差完全取决于 (partial _t u_1 +{{textrm{div},}}_textbf{x} textbf{u}_2/)的光滑度,即强制项 (f=(partial _t-Delta _x)u/)的光滑度。数值结果表明收敛速度明显提高。
{"title":"Improved rates for a space–time FOSLS of parabolic PDEs","authors":"Gregor Gantner, Rob Stevenson","doi":"10.1007/s00211-023-01387-3","DOIUrl":"https://doi.org/10.1007/s00211-023-01387-3","url":null,"abstract":"<p>We consider the first-order system space–time formulation of the heat equation introduced by Bochev and Gunzburger (in: Bochev and Gunzburger (eds) Applied mathematical sciences, vol 166, Springer, New York, 2009), and analyzed by Führer and Karkulik (Comput Math Appl 92:27–36, 2021) and Gantner and Stevenson (ESAIM Math Model Numer Anal 55(1):283–299 2021), with solution components <span>((u_1,textbf{u}_2)=(u,-nabla _textbf{x} u))</span>. The corresponding operator is boundedly invertible between a Hilbert space <i>U</i> and a Cartesian product of <span>(L_2)</span>-type spaces, which facilitates easy first-order system least-squares (FOSLS) discretizations. Besides <span>(L_2)</span>-norms of <span>(nabla _textbf{x} u_1)</span> and <span>(textbf{u}_2)</span>, the (graph) norm of <i>U</i> contains the <span>(L_2)</span>-norm of <span>(partial _t u_1 +{{,textrm{div},}}_textbf{x} textbf{u}_2)</span>. When applying standard finite elements w.r.t. simplicial partitions of the space–time cylinder, estimates of the approximation error w.r.t. the latter norm require higher-order smoothness of <span>(textbf{u}_2)</span>. In experiments for both uniform and adaptively refined partitions, this manifested itself in disappointingly low convergence rates for non-smooth solutions <i>u</i>. In this paper, we construct finite element spaces w.r.t. prismatic partitions. They come with a quasi-interpolant that satisfies a near commuting diagram in the sense that, apart from some harmless term, the aforementioned error depends exclusively on the smoothness of <span>(partial _t u_1 +{{,textrm{div},}}_textbf{x} textbf{u}_2)</span>, i.e., of the forcing term <span>(f=(partial _t-Delta _x)u)</span>. Numerical results show significantly improved convergence rates.</p>","PeriodicalId":49733,"journal":{"name":"Numerische Mathematik","volume":null,"pages":null},"PeriodicalIF":2.1,"publicationDate":"2023-12-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139053881","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"数学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-12-18DOI: 10.1007/s00211-023-01386-4
Philip Freese, Moritz Hauck, Tim Keil, Daniel Peterseim
This paper presents a novel multi-scale method for elliptic partial differential equations with arbitrarily rough coefficients. In the spirit of numerical homogenization, the method constructs problem-adapted ansatz spaces with uniform algebraic approximation rates. Localized basis functions with the same super-exponential localization properties as the recently proposed Super-Localized Orthogonal Decomposition enable an efficient implementation. The method’s basis stability is enforced using a partition of unity approach. A natural extension to higher order is presented, resulting in higher approximation rates and enhanced localization properties. We perform a rigorous a priori and a posteriori error analysis and confirm our theoretical findings in a series of numerical experiments. In particular, we demonstrate the method’s applicability for challenging high-contrast channeled coefficients.
{"title":"A super-localized generalized finite element method","authors":"Philip Freese, Moritz Hauck, Tim Keil, Daniel Peterseim","doi":"10.1007/s00211-023-01386-4","DOIUrl":"https://doi.org/10.1007/s00211-023-01386-4","url":null,"abstract":"<p>This paper presents a novel multi-scale method for elliptic partial differential equations with arbitrarily rough coefficients. In the spirit of numerical homogenization, the method constructs problem-adapted ansatz spaces with uniform algebraic approximation rates. Localized basis functions with the same super-exponential localization properties as the recently proposed Super-Localized Orthogonal Decomposition enable an efficient implementation. The method’s basis stability is enforced using a partition of unity approach. A natural extension to higher order is presented, resulting in higher approximation rates and enhanced localization properties. We perform a rigorous a priori and a posteriori error analysis and confirm our theoretical findings in a series of numerical experiments. In particular, we demonstrate the method’s applicability for challenging high-contrast channeled coefficients.</p>","PeriodicalId":49733,"journal":{"name":"Numerische Mathematik","volume":null,"pages":null},"PeriodicalIF":2.1,"publicationDate":"2023-12-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"138716249","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"数学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-12-18DOI: 10.1007/s00211-023-01389-1
Rommel R. Real
We consider the Landweber iteration for solving linear as well as nonlinear inverse problems in Banach spaces. Based on the discrepancy principle, we propose a heuristic parameter choice rule for choosing the regularization parameter which does not require the information on the noise level, so it is purely data-driven. According to a famous veto, convergence in the worst-case scenario cannot be expected in general. However, by imposing certain conditions on the noisy data, we establish a new convergence result which, in addition, requires neither the Gâteaux differentiability of the forward operator nor the reflexivity of the image space. Therefore, we also expand the applied range of the Landweber iteration to cover non-smooth ill-posed inverse problems and to handle the situation that the data is contaminated by various types of noise. Numerical simulations are also reported.
{"title":"Hanke–Raus rule for Landweber iteration in Banach spaces","authors":"Rommel R. Real","doi":"10.1007/s00211-023-01389-1","DOIUrl":"https://doi.org/10.1007/s00211-023-01389-1","url":null,"abstract":"<p>We consider the Landweber iteration for solving linear as well as nonlinear inverse problems in Banach spaces. Based on the discrepancy principle, we propose a heuristic parameter choice rule for choosing the regularization parameter which does not require the information on the noise level, so it is purely data-driven. According to a famous veto, convergence in the worst-case scenario cannot be expected in general. However, by imposing certain conditions on the noisy data, we establish a new convergence result which, in addition, requires neither the Gâteaux differentiability of the forward operator nor the reflexivity of the image space. Therefore, we also expand the applied range of the Landweber iteration to cover non-smooth ill-posed inverse problems and to handle the situation that the data is contaminated by various types of noise. Numerical simulations are also reported.</p>","PeriodicalId":49733,"journal":{"name":"Numerische Mathematik","volume":null,"pages":null},"PeriodicalIF":2.1,"publicationDate":"2023-12-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"138744165","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"数学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-12-13DOI: 10.1007/s00211-023-01388-2
Aekta Aggarwal, Helge Holden, Ganesh Vaidya
In this article, we discuss the error analysis for a certain class of monotone finite volume schemes approximating nonlocal scalar conservation laws, modeling traffic flow and crowd dynamics, without any additional assumptions on monotonicity or linearity of the kernel (mu ) or the flux f. We first prove a novel Kuznetsov-type lemma for this class of PDEs and thereby show that the finite volume approximations converge to the entropy solution at the rate of (sqrt{Delta t}) in (L^1(mathbb {R})). To the best of our knowledge, this is the first proof of any type of convergence rate for this class of conservation laws. We also present numerical experiments to illustrate this result.
{"title":"On the accuracy of the finite volume approximations to nonlocal conservation laws","authors":"Aekta Aggarwal, Helge Holden, Ganesh Vaidya","doi":"10.1007/s00211-023-01388-2","DOIUrl":"https://doi.org/10.1007/s00211-023-01388-2","url":null,"abstract":"<p>In this article, we discuss the error analysis for a certain class of monotone finite volume schemes approximating nonlocal scalar conservation laws, modeling traffic flow and crowd dynamics, without any additional assumptions on monotonicity or linearity of the kernel <span>(mu )</span> or the flux <i>f</i>. We first prove a novel Kuznetsov-type lemma for this class of PDEs and thereby show that the finite volume approximations converge to the entropy solution at the rate of <span>(sqrt{Delta t})</span> in <span>(L^1(mathbb {R}))</span>. To the best of our knowledge, this is the first proof of any type of convergence rate for this class of conservation laws. We also present numerical experiments to illustrate this result.</p>","PeriodicalId":49733,"journal":{"name":"Numerische Mathematik","volume":null,"pages":null},"PeriodicalIF":2.1,"publicationDate":"2023-12-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"138683261","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"数学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-12-11DOI: 10.1007/s00211-023-01376-6
Moulay Abdellah Chkifa
Given a positive integer d and ({{varvec{a}}}_{1},dots ,{{varvec{a}}}_{r}) a family of vectors in ({{mathbb {R}}}^d), ({k_1{{varvec{a}}}_{1}+dots +k_r{{varvec{a}}}_{r}: k_1,dots ,k_r in {{mathbb {Z}}}}subset {{mathbb {R}}}^d) is the so-called lattice generated by the family. In high dimensional integration, prescribed lattices are used for constructing reliable quadrature schemes. The quadrature points are the lattice points lying on the integration domain, typically the unit hypercube ([0,1)^d) or a rescaled shifted hypercube. It is crucial to have a cost-effective method for enumerating lattice points within such domains. Undeniably, the lack of such fast enumeration procedures hinders the applicability of lattice rules. Existing enumeration procedures exploit intrinsic properties of the lattice at hand, such as periodicity, orthogonality, recurrences, etc. In this paper, we unveil a general-purpose fast lattice enumeration algorithm based on linear programming (named FLE-LP).
给定一个正整数 d 和 ({{varvec{a}}}_{1},dots ,{{{varvec{a}}}_{r}) 中的一个向量族, ({k_1{{varvec{a}}}_{1}+dots +k_r{{varvec{a}}}_{r}:k_1,dots ,k_r in {{mathbb {Z}}} 子集 {{mathbb {R}}}^d) 是由族产生的所谓晶格。在高维积分中,规定网格用于构建可靠的正交方案。正交点是位于积分域上的网格点,通常是单位超立方体(([0,1)^d)或重比例移位超立方体。拥有一种经济有效的方法来枚举这些域内的网格点至关重要。不可否认,缺乏这种快速枚举程序阻碍了网格规则的应用。现有的枚举程序利用的是网格的内在属性,如周期性、正交性、递归性等。本文揭示了一种基于线性规划的通用快速网格枚举算法(命名为 FLE-LP)。
{"title":"Lattice enumeration via linear programming","authors":"Moulay Abdellah Chkifa","doi":"10.1007/s00211-023-01376-6","DOIUrl":"https://doi.org/10.1007/s00211-023-01376-6","url":null,"abstract":"<p>Given a positive integer <i>d</i> and <span>({{varvec{a}}}_{1},dots ,{{varvec{a}}}_{r})</span> a family of vectors in <span>({{mathbb {R}}}^d)</span>, <span>({k_1{{varvec{a}}}_{1}+dots +k_r{{varvec{a}}}_{r}: k_1,dots ,k_r in {{mathbb {Z}}}}subset {{mathbb {R}}}^d)</span> is the so-called lattice generated by the family. In high dimensional integration, prescribed lattices are used for constructing reliable quadrature schemes. The quadrature points are the lattice points lying on the integration domain, typically the unit hypercube <span>([0,1)^d)</span> or a rescaled shifted hypercube. It is crucial to have a cost-effective method for enumerating lattice points within such domains. Undeniably, the lack of such fast enumeration procedures hinders the applicability of lattice rules. Existing enumeration procedures exploit intrinsic properties of the lattice at hand, such as periodicity, orthogonality, recurrences, etc. In this paper, we unveil a general-purpose fast lattice enumeration algorithm based on linear programming (named <b>FLE-LP</b>).</p>","PeriodicalId":49733,"journal":{"name":"Numerische Mathematik","volume":null,"pages":null},"PeriodicalIF":2.1,"publicationDate":"2023-12-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"138572065","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"数学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-12-07DOI: 10.1007/s00211-023-01385-5
Dietmar Gallistl, Ngoc Tien Tran
This paper analyzes a regularization scheme of the Monge–Ampère equation by uniformly elliptic Hamilton–Jacobi–Bellman equations. The main tools are stability estimates in the (L^infty ) norm from the theory of viscosity solutions which are independent of the regularization parameter (varepsilon ). They allow for the uniform convergence of the solution (u_varepsilon ) to the regularized problem towards the Alexandrov solution u to the Monge–Ampère equation for any nonnegative (L^n) right-hand side and continuous Dirichlet data. The main application are guaranteed a posteriori error bounds in the (L^infty ) norm for continuously differentiable finite element approximations of u or (u_varepsilon ).
{"title":"Stability and guaranteed error control of approximations to the Monge–Ampère equation","authors":"Dietmar Gallistl, Ngoc Tien Tran","doi":"10.1007/s00211-023-01385-5","DOIUrl":"https://doi.org/10.1007/s00211-023-01385-5","url":null,"abstract":"<p>This paper analyzes a regularization scheme of the Monge–Ampère equation by uniformly elliptic Hamilton–Jacobi–Bellman equations. The main tools are stability estimates in the <span>(L^infty )</span> norm from the theory of viscosity solutions which are independent of the regularization parameter <span>(varepsilon )</span>. They allow for the uniform convergence of the solution <span>(u_varepsilon )</span> to the regularized problem towards the Alexandrov solution <i>u</i> to the Monge–Ampère equation for any nonnegative <span>(L^n)</span> right-hand side and continuous Dirichlet data. The main application are guaranteed a posteriori error bounds in the <span>(L^infty )</span> norm for continuously differentiable finite element approximations of <i>u</i> or <span>(u_varepsilon )</span>.\u0000</p>","PeriodicalId":49733,"journal":{"name":"Numerische Mathematik","volume":null,"pages":null},"PeriodicalIF":2.1,"publicationDate":"2023-12-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"138546841","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"数学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-11-30DOI: 10.1007/s00211-023-01381-9
Snorre H. Christiansen, Jay Gopalakrishnan, Johnny Guzmán, Kaibo Hu
We construct conforming finite element elasticity complexes on the Alfeld splits of tetrahedra. The complex consists of vector fields and symmetric tensor fields, interlinked via the linearized deformation operator, the linearized curvature operator, and the divergence operator, respectively. The construction is based on an algebraic machinery that derives the elasticity complex from de Rham complexes, and smoother finite element differential forms.
{"title":"A discrete elasticity complex on three-dimensional Alfeld splits","authors":"Snorre H. Christiansen, Jay Gopalakrishnan, Johnny Guzmán, Kaibo Hu","doi":"10.1007/s00211-023-01381-9","DOIUrl":"https://doi.org/10.1007/s00211-023-01381-9","url":null,"abstract":"<p>We construct conforming finite element elasticity complexes on the Alfeld splits of tetrahedra. The complex consists of vector fields and symmetric tensor fields, interlinked via the linearized deformation operator, the linearized curvature operator, and the divergence operator, respectively. The construction is based on an algebraic machinery that derives the elasticity complex from de Rham complexes, and smoother finite element differential forms.\u0000</p>","PeriodicalId":49733,"journal":{"name":"Numerische Mathematik","volume":null,"pages":null},"PeriodicalIF":2.1,"publicationDate":"2023-11-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"138531609","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"数学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-11-23DOI: 10.1007/s00211-023-01384-6
Yuanyuan Li, Shuai Lu, Peter Mathé, Sergei V. Pereverzev
We investigate the use of two-layer networks with the rectified power unit, which is called the (text {ReLU}^k) activation function, for function and derivative approximation. By extending and calibrating the corresponding Barron space, we show that two-layer networks with the (text {ReLU}^k) activation function are well-designed to simultaneously approximate an unknown function and its derivatives. When the measurement is noisy, we propose a Tikhonov type regularization method, and provide error bounds when the regularization parameter is chosen appropriately. Several numerical examples support the efficiency of the proposed approach.
{"title":"Two-layer networks with the $$text {ReLU}^k$$ activation function: Barron spaces and derivative approximation","authors":"Yuanyuan Li, Shuai Lu, Peter Mathé, Sergei V. Pereverzev","doi":"10.1007/s00211-023-01384-6","DOIUrl":"https://doi.org/10.1007/s00211-023-01384-6","url":null,"abstract":"<p>We investigate the use of two-layer networks with the rectified power unit, which is called the <span>(text {ReLU}^k)</span> activation function, for function and derivative approximation. By extending and calibrating the corresponding Barron space, we show that two-layer networks with the <span>(text {ReLU}^k)</span> activation function are well-designed to simultaneously approximate an unknown function and its derivatives. When the measurement is noisy, we propose a Tikhonov type regularization method, and provide error bounds when the regularization parameter is chosen appropriately. Several numerical examples support the efficiency of the proposed approach.</p>","PeriodicalId":49733,"journal":{"name":"Numerische Mathematik","volume":null,"pages":null},"PeriodicalIF":2.1,"publicationDate":"2023-11-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"138531594","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"数学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}