An artificial neural network approximation for Cauchy inverse problems

A novel artificial neural network method is proposed for solving Cauchy inverse problems. It allows multiple hidden layers with arbitrary width and depth, which theoretically yields better approximations to the inverse problems. In this research, the existence and convergence are shown to establish the well-posedness of neural network method for Cauchy inverse problems, and various numerical examples are presented to illustrate its accuracy and stability. The numerical examples are from different points of view, including time-dependent and time-independent cases, high spatial dimension cases up to 8D, and cases with noisy boundary data and singular computational domain. Moreover, numerical results also show that neural networks with wider and deeper hidden layers could lead to better approximation for Cauchy inverse problems.

[1]  George Em Karniadakis,et al.  fPINNs: Fractional Physics-Informed Neural Networks , 2018, SIAM J. Sci. Comput..

[2]  Kurt Hornik,et al.  Multilayer feedforward networks are universal approximators , 1989, Neural Networks.

[3]  A. Nachaoui,et al.  Convergence of an Alternating Method to Solve the Cauchy Problem for Poisson's Equation , 2002 .

[4]  Bangti Jin,et al.  Hierarchical Bayesian inference for Ill-posed problems via variational method , 2010, J. Comput. Phys..

[5]  Alaeddin Malek,et al.  Numerical solution for high order differential equations using a hybrid neural network - Optimization method , 2006, Appl. Math. Comput..

[6]  A. Nachaoui,et al.  Convergence analysis for finite element approximation to an inverse Cauchy problem , 2006 .

[7]  Lars Eldén,et al.  Approximations for a Cauchy problem for the heat equation , 1987 .

[8]  Kurt Hornik,et al.  Approximation capabilities of multilayer feedforward networks , 1991, Neural Networks.

[9]  Daniel A. White,et al.  Multiscale topology optimization using neural network surrogate models , 2019, Computer Methods in Applied Mechanics and Engineering.

[10]  Masahiro Yamamoto,et al.  Tikhonov Regularization by a Reproducing Kernel Hilbert Space for the Cauchy Problem for an Elliptic Equation , 2008, SIAM J. Sci. Comput..

[11]  Siddhartha Mishra,et al.  A machine learning framework for data driven acceleration of computations of differential equations , 2018, ArXiv.

[12]  Dimitrios I. Fotiadis,et al.  Artificial neural networks for solving ordinary and partial differential equations , 1997, IEEE Trans. Neural Networks.

[13]  Fredrik Berntsson,et al.  Iterative Tikhonov regularization for the Cauchy problem for the Helmholtz equation , 2017, Comput. Math. Appl..

[14]  T. Wei,et al.  An inverse time-dependent source problem for a time-space fractional diffusion equation , 2018, Appl. Math. Comput..

[15]  Derek B. Ingham,et al.  An alternating iterative algorithm for the Cauchy problem associated to the Helmholtz equation , 2003 .

[16]  Bangti Jin,et al.  Augmented Tikhonov regularization , 2009 .

[17]  Paris Perdikaris,et al.  Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations , 2019, J. Comput. Phys..

[18]  Hans-Jürgen Reinhardt,et al.  Stability and Regularization of a Discrete Approximation to the Cauchy Problem for Laplace's Equation , 1999 .

[19]  Lars Eldén,et al.  Solving a Cauchy problem for a 3D elliptic PDE with variable coefficients by a quasi-boundary-value method , 2014 .

[20]  Jim Douglas,et al.  The Cauchy Problem for the Heat Equation , 1967 .

[21]  Matthias Troyer,et al.  Solving the quantum many-body problem with artificial neural networks , 2016, Science.

[22]  Liang Yan,et al.  Adaptive multi-fidelity polynomial chaos approach to Bayesian inference in inverse problems , 2018, J. Comput. Phys..

[23]  Lucie P. Aarts,et al.  Neural Network Method for Solving Partial Differential Equations , 2001, Neural Processing Letters.

[24]  George Cybenko,et al.  Approximation by superpositions of a sigmoidal function , 1989, Math. Control. Signals Syst..

[25]  A. Nachaoui.  Numerical linear algebra for reconstruction inverse problems , 2004 .

[26]  S. Kamin,et al.  Uniqueness of solutions of the Cauchy problem for parabolic equations degenerating at infinity , 2000 .

[27]  Kailiang Wu,et al.  Data Driven Governing Equations Approximation Using Deep Neural Networks , 2018, J. Comput. Phys..

[28]  George Em Karniadakis,et al.  Quantifying total uncertainty in physics-informed neural networks for solving forward and inverse stochastic problems , 2018, J. Comput. Phys..

[29]  Xiaoliang Cheng,et al.  A dynamical regularization algorithm for solving inverse source problems of elliptic partial differential equations , 2018 .

[30]  Laurent Bourgeois,et al.  A mixed formulation of quasi-reversibility to solve the Cauchy problem for Laplace's equation , 2005 .

[31]  Ting Wei,et al.  Backus-Gilbert algorithm for the Cauchy problem of the Laplace equation , 2001 .

[32]  Masahiro Yamamoto,et al.  Initial value/boundary value problems for fractional diffusion-wave equations and applications to some inverse problems , 2011 .

[33]  Michael V. Klibanov,et al.  A computational quasi-reversiblility method for Cauchy problems for Laplace's equation , 1991 .

[34]  Dimitris G. Papageorgiou,et al.  Neural-network methods for boundary value problems with irregular boundaries , 2000, IEEE Trans. Neural Networks Learn. Syst..

[35]  T. Wei,et al.  Two regularization methods for the Cauchy problems of the Helmholtz equation , 2010 .

[36]  Mohamed Jaoua,et al.  Solution of the Cauchy problem using iterated Tikhonov regularization , 2001 .

[37]  Long Chen,et al.  Maximum Principle Based Algorithms for Deep Learning , 2017, J. Mach. Learn. Res..

[38]  Bin Dong,et al.  PDE-Net 2.0: Learning PDEs from Data with A Numeric-Symbolic Hybrid Deep Network , 2018, J. Comput. Phys..

[39]  W. Han,et al.  A novel coupled complex boundary method for solving inverse source problems , 2014 .

[40]  V. A. Kozlov,et al.  An accelerated alternating procedure for the Cauchy problem for the Helmholtz equation , 2014, Comput. Math. Appl..

[41]  Justin A. Sirignano,et al.  DGM: A deep learning algorithm for solving partial differential equations , 2017, J. Comput. Phys..

[42]  Y. G. Chen,et al.  A variational-type method of fundamental solutions for a Cauchy problem of Laplace’s equation , 2013 .

[43]  D. Aronson.  Uniqueness of solutions of the Cauchy problem for parabolic equations*1 , 1966 .