A Hybrid Iterative Method for Elliptic Variational Inequalities of the Second Kind

Abstract

A hybrid iterative method is proposed for numerically solving the elliptic variational inequality (EVI) of the second kind, through combining the regularized semi-smooth Newton method and the Int-Deep method. The convergence rate analysis and numerical examples on contact problems show this algorithm converges rapidly and is efficient for solving EVIs.

Keywords

Deep learning / Semi-smooth Newton method / Elliptic variational inequality (EVI) / Convergence rate analysis

References

[1] AdamsRASobolev Spaces, 1975New York/LondonAcademic Press

[2] BarronAR. Universal approximation bounds for superpositions of a sigmoidal function. IEEE Trans. Inform. Theory, 1993, 393930-945.

[3] BottouL, CurtisFE, NocedalJ. Optimization methods for large-scale machine learning. SIAM Rev., 2018, 602223-311.

[4] BrennerSC. Korn’s inequalities for piecewise $H^1$ vector fields. Math. Comp., 2004, 732471067-1087.

[5] BrezziF, HagerWW, RaviartP-A. Error estimates for the finite element solution of variational inequalities. Numer. Math., 1977, 284431-443.

[6] Brezzi, F., Hager, W.W., Raviart, P.-A.: Error estimates for the finite element solution of variational inequalities. II. Mixed methods. Numer. Math. 31(1), 1–16 (1978)

[7] ChenF, HuangJ, WangC, YangH. Friedrichs learning: weak solutions of partial differential equations via deep learning. SIAM J. Sci. Comput., 2023, 4531271-1299.

[8] Chen, J., Chi, X., E, W., Yang, Z.: Bridging traditional and machine learning-based algorithms for solving PDEs: the random feature method. J. Mach. Learn 1(3), 268–298 (2022)

[9] ChengX-L, HanW. Inexact Uzawa algorithms for variational inequalities of the second kind. Comput. Methods Appl. Mech. Engrg., 2003, 19211/121451-1462.

[10] ChengX-L, ShenX, WangX, LiangK. A deep neural network-based method for solving obstacle problems. Nonlinear Anal. Real World Appl., 2023, 72. 103864

[11] CiarletPGThe Finite Element Method for Elliptic Problems, 1978Amsterdam/New York/OxfordNorth-Holland Publishing Co.

[12] CombettesPL, WajsVR. Signal recovery by proximal forward-backward splitting. Multiscale Model. Simul., 2005, 441168-1200.

[13] Cuomo, S., Schiano Di Cola, V., Giampaolo, F., Rozza, G., Raissi, M., Piccialli, F.: Scientific machine learning through physics-informed neural networks: where we are and what’s next. J. Sci. Comput. 92(3), 88 (2022)

[14] DuvautG, LionsJ-LInequalities in Mechanics and Physics, 1976Berlin/New YorkSpringer.

[15] E, W.N. Machine learning and computational mathematics. Commun. Comput. Phys. 28(5), 1639–1670 (2020)

[16] E, W.N., Han, J., Jentzen, A.: Algorithms for solving high dimensional PDEs: from nonlinear Monte Carlo to machine learning. Nonlinearity 35(1), 278–310 (2022)

[17] E, W. N., Ma, C., Wu, L.: A priori estimates of the population risk for two-layer neural networks. Commun. Math. Sci. 17(5), 1407–1425 (2019)

[18] E, W. N., Yu, B.: The deep Ritz method: a deep learning-based numerical algorithm for solving variational problems. Commun. Math. Stat. 6(1), 1–12 (2018)

[19] FanJ-Y, YuanY-X. On the quadratic convergence of the Levenberg-Marquardt method without nonsingularity assumption. Computing, 2005, 74123-39.

[20] FengF, HanW, HuangJ. Virtual element method for an elliptic hemivariational inequality with applications to contact mechanics. J. Sci. Comput., 2019, 8132388-2412.

[21] GaoZ, YanL, ZhouT. Failure-informed adaptive sampling for PINNs. SIAM J. Sci. Comput., 2023, 4541971-1994.

[22] GlowinskiRNumerical Methods for Nonlinear Variational Problems, 1984New YorkSpringer.

[23] GlowinskiR, LionsJ-L, TrémolièresRNumerical Analysis of Variational Inequalities, 1981Amsterdam/New YorkNorth-Holland Publishing Co.

[24] GoodfellowI, BengioY, CourvilleADeep Learning, 2016CambridgeMIT Press

[25] HanW, WangL-H. Nonconforming finite element analysis for a plate contact problem. SIAM J. Numer. Anal., 2002, 4051683-1697.

[26] He, K., Zhang, X., Ren, S., Sun, J.: Deep residual learning for image recognition. In: The 29th IEEE Conference on Computer Vision and Pattern Recognition, IEEE, Las Vegas, 27–30 June, 2016 (2016)

[27] HornikK, StinchcombeM, WhiteH. Multilayer feedforward networks are universal approximators. Neural Netw., 1989, 25359-366.

[28] HuangJ, WangH, YangH. Int-Deep: a deep learning initialized iterative method for nonlinear problems. J. Comput. Phys., 2020, 419. 109675

[29] HuangJ, WangH, ZhouT. An augmented Lagrangian deep learning method for variational problems with essential boundary conditions. Commun. Comput. Phys., 2022, 313966-986.

[30] Hüeber, S., Wohlmuth, B.I.: A primal-dual active set strategy for non-linear multibody contact problems. Comput. Methods Appl. Mech. Engrg. 194(27/28/29), 3147–3166 (2005)

[31] KinderlehrerD, StampacchiaGAn Introduction to Variational Inequalities and Their Applications, 1980PhiladelphiaSociety for Industrial and Applied Mathematics

[32] Kingma, D.P., Ba, J.: Adam: a method for stochastic optimization. In: The 3rd International Conference on Learning Representations, ICLR, San Diego, 7–9 May, 2015 (2015)

[33] LionsJ-L, StampacchiaG. Variational inequalities. Comm. Pure Appl. Math., 1967, 20: 493-519.

[34] PangJ-S, QiLQ. Nonsmooth equations: motivation and algorithms. SIAM J. Optim., 1993, 33443-465.

[35] QiLQ. Convergence analysis of some algorithms for solving nonsmooth equations. Math. Oper. Res., 1993, 181227-244.

[36] RaissiM, PerdikarisP, KarniadakisGE. Physics-informed neural networks: a deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations. J. Comput. Phys., 2019, 378: 686-707.

[37] ShenZ, YangH, ZhangS. Optimal approximation rate of ReLU networks in terms of width and depth. J. Math. Pures Appl., 2022, 157: 101-135.

[38] SiegelJW, XuJ. High-order approximation rates for shallow neural networks with cosine and ${\rm ReLU}^k$ activation functions. Appl. Comput. Harmon. Anal., 2022, 58: 1-26.

[39] StadlerG. Semismooth Newton and augmented Lagrangian methods for a simplified friction problem. SIAM J. Optim., 2004, 15139-62.

[40] WangF, HanW, ChengX-L. Discontinuous Galerkin methods for solving elliptic variational inequalities. SIAM J. Numer. Anal., 2010, 482708-733.

[41] WangS, TengY, PerdikarisP. Understanding and mitigating gradient flow pathologies in physics-informed neural networks. SIAM J. Sci. Comput., 2021, 4353055-3081.

[42] XiaoX, LiY, WenZ, ZhangL. A regularized semi-smooth newton method with projection steps for composite convex programs. J. Sci. Comput., 2018, 761364-389.

[43] XuZ-QJ, ZhangY, LuoT, XiaoY, MaZ. Frequency principle: Fourier analysis sheds light on deep neural networks. Commun. Comput. Phys., 2020, 2851746-1767.

[44] ZangY, BaoG, YeX, ZhouH. Weak adversarial networks for high-dimensional partial differential equations. J. Comput. Phys., 2020, 411. 109409

Funding

National Natural Science Foundation of China(12071289)

RIGHTS & PERMISSIONS

Shanghai University