Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
112 tokens/sec
GPT-4o
8 tokens/sec
Gemini 2.5 Pro Pro
47 tokens/sec
o3 Pro
5 tokens/sec
GPT-4.1 Pro
38 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

Solving Elliptic Optimal Control Problems via Neural Networks and Optimality System (2308.11925v2)

Published 23 Aug 2023 in math.OC, cs.LG, cs.NA, and math.NA

Abstract: In this work, we investigate a neural network based solver for optimal control problems (without / with box constraint) for linear and semilinear second-order elliptic problems. It utilizes a coupled system derived from the first-order optimality system of the optimal control problem, and employs deep neural networks to represent the solutions to the reduced system. We present an error analysis of the scheme, and provide $L2(\Omega)$ error bounds on the state, control and adjoint in terms of neural network parameters (e.g., depth, width, and parameter bounds) and the numbers of sampling points. The main tools in the analysis include offset Rademacher complexity and boundedness and Lipschitz continuity of neural network functions. We present several numerical examples to illustrate the method and compare it with two existing ones.

Definition Search Book Streamline Icon: https://streamlinehq.com
References (53)
  1. M. Anthony and P. L. Bartlett. Neural Network Learning: Theoretical Foundations. Cambridge University Press, Cambridge, 1999.
  2. Physics-informed neural networks for PDE-constrained optimization and control. Preprint, arXiv:2205.03377, 2022.
  3. P. L. Bartlett and S. Mendelson. Rademacher and Gaussian complexities: risk bounds and structural results. J. Mach. Learn. Res., 3:463–482, 2002.
  4. Automatic differentiation in machine learning: a survey. J. Mach. Learn. Res., 18:1–43, 2018.
  5. An overview on deep learning-based approximation methods for partial differential equations. Discrete Contin. Dyn. Syst. Ser. B, 28(6):3697–3746, 2023.
  6. M. Berggren. Approximations of very weak solutions to boundary-value problems. SIAM J. Numer. Anal., 42(2):860–877, 2004.
  7. D. P. Bertsekas. Constrained Optimization and Lagrange Multiplier Methods. Academic Press, Inc., New York-London, 1982.
  8. Data-driven initialization of deep learning solvers for Hamilton-Jacobi-Bellman PDEs. IFAC, 55(30):168–173, 2022.
  9. A limited memory algorithm for bound constrained optimization. SIAM J. Sci. Comput., 16(5):1190–1208, 1995.
  10. Physics-informed neural networks (PINNs) for fluid mechanics: a review. Acta Mech. Sin., 37(12):1727–1738, 2021.
  11. Least-squares ReLU neural network (LSNN) method for linear advection-reaction equation. J. Comput. Phys., 443:110514, 17, 2021.
  12. J. Céa. Conception optimale ou identification de formes: calcul rapide de la dérivée directionnelle de la fonction coût. RAIRO Modél. Math. Anal. Numér., 20(3):371–402, 1986.
  13. F. Cucker and S. Smale. On the mathematical foundations of learning. Bull. Amer. Math. Soc. (N.S.), 39(1):1–49, 2002.
  14. An extended physics informed neural network for preliminary analysis of parametric optimal control problems. Comput. Math. Appl., 143:383–396, 2023.
  15. Fast excess risk rates via offset Rademacher complexity. In Proceedings of the 40th International Conference on Machine Learning, pages 8697–8716, 2023.
  16. W. E and B. Yu. The deep Ritz method: a deep learning-based numerical algorithm for solving variational problems. Commun. Math. Stat., 6(1):1–12, 2018.
  17. R. S. Falk. Approximation of a class of optimal control problems with order of convergence estimates. J. Math. Anal. Appl., 44:28–47, 1973.
  18. I. Gühring and M. Raslan. Approximation rates for neural networks with encodable weights in smoothness spaces. Neural Networks, 134:107–130, 2021.
  19. A physics-informed deep learning framework for inversion and surrogate modeling in solid mechanics. Comput. Methods Appl. Mech. Engrg., 379:Paper No. 113741, 22, 2021.
  20. W. Hoeffding. Probability inequalities for sums of bounded random variables. J. Amer. Statist. Assoc., 58:13–30, 1963.
  21. D. Hoff and H. Wendland. A meshfree method for a PDE-constrained optimization problem. SIAM J. Numer. Anal., 59(4):1896–1917, 2021.
  22. Solving elliptic problems with singular sources using singularity splitting deep Ritz method. SIAM J. Sci. Comput., 45(4):A2043–A2074, 2023.
  23. Solving Poisson problems in polygonal domains with singularity enriched physics informed neural networks. SIAM J. Sci. Comput., page in press. arXiv:2308.16429, 2024.
  24. A. Jameson. Aerodynamic design via control theory. In Recent Advances in Computational Fluid Dynamics (Princeton, NJ, 1988), volume 43 of Lecture Notes in Engrg., pages 377–401. Springer, Berlin, 1989.
  25. A rate of convergence of physics informed neural networks for the linear second order elliptic pdes. Comm. Comput. Phys., 31(4):1272–1295, 2022.
  26. Error analysis of deep Ritz methods for elliptic equations. Anal. Appl. (Singap.), 22(1):57–87, 2024.
  27. Physics-informed machine learning. Nature Rev. Phys., 3:422–440, 2021.
  28. D. P. Kingma and J. Ba. Adam: A method for stochastic optimization. In 3rd International Conference for Learning Representations, San Diego, 2015.
  29. Characterizing possible failure modes in physics-informed neural networks. In Advances in Neural Information Processing Systems 34, pages 26548–26560, 2021.
  30. Artificial neural networks for solving ordinary and partial differential equations. IEEE Trans. Neural Netw., 9(5):987–1000, 1998.
  31. Neural-network methods for boundary value problems with irregular boundaries. IEEE Trans. Neural Netw., 11(5):1041–1049, 2000.
  32. H. Lee and I. S. Kang. Neural algorithm for solving differential equations. J. Comput. Phys., 91(1):110–131, 1990.
  33. Learning with square loss: localization through offset Rademacher complexity. In Proceedings of The 28th Conference on Learning Theory, PMLR 40, pages 1260–1285, 2015.
  34. J.-L. Lions. Optimal Control of Systems Governed by Partial Differential Equations. Springer-Verlag, New York-Berlin, 1971.
  35. Physics-informed neural networks with hard constraints for inverse design. SIAM J. Sci. Comput., 43(6):B1105–B1132, 2021.
  36. Machine learning for elliptic PDEs: Fast rate generalization bound, neural scaling law and minimax optimality. Preprint, arXiv:2110.06897, 2021.
  37. A priori generalization error analysis of the deep Ritz method for solving high dimensional elliptic partial differential equations. In Conference in Learning Theory, pages 3196–3241, 2021.
  38. Optimal Control of Partial Differential Equations — Analysis, Approximation, and Applications. Springer, Cham, 2021.
  39. SympOCnet: solving optimal control problems with applications to high-dimensional multiagent path planning problems. SIAM J. Sci. Comput., 44(6):B1341–B1368, 2022.
  40. S. Mishra and R. Molinaro. Estimates on the generalization error of physics-informed neural networks for approximating PDEs. IMA J. Numer. Anal., 43(1):1–43, 2023.
  41. S. Mowlavi and S. Nabi. Optimal control of PDEs using physics-informed neural networks. J. Comput. Phys., 473:111731, 22, 2023.
  42. Adaptive deep learning for high-dimensional Hamilton-Jacobi-Bellman equations. SIAM J. Sci. Comput., 43(2):A1221–A1247, 2021.
  43. Physics-informed neural networks: a deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations. J. Comput. Phys., 378:686–707, 2019.
  44. On the convergence of physics informed neural networks for linear second-order elliptic and parabolic type PDEs. Commun. Comput. Phys., 28(5):2042–2074, 2020.
  45. Greedy training algorithms for neural networks and applications to PDEs. J. Comput. Phys., 484:112084, 27, 2023.
  46. J. Sirignano and K. Spiliopoulos. DGM: A deep learning algorithm for solving partial differential equations. J. Comput. Phys., 375:1339–1364, 2018.
  47. The ADMM-PINNs algorithmic framework for nonsmooth PDE-constrained optimization: a deep learning approach. Preprint, arXiv:2302.08309, 2023.
  48. Deep learning methods for partial differential equations and related parameter identification problems. Inverse Problems, 39(10):103001, 75, 2023.
  49. F. Tröltzsch. Optimal Control of Partial Differential Equations. AMS, Providence, RI, 2010.
  50. When and why PINNs fail to train: a neural tangent kernel perspective. J. Comput. Phys., 449:110768, 28, 2022.
  51. AONN: an adjoint-oriented neural network method for all-at-once solutions of parametric optimal control problems. SIAM J. Sci. Comput., 46(1):C127–C153, 2024.
  52. Weak adversarial networks for high-dimensional partial differential equations. J. Comput. Phys., 411:109409, 14, 2020.
  53. Y. Zhao and J. Han. Offline supervised learning v.s. online direct policy optimization: A comparative study and a unified training paradigm for neural network-based optimal feedback control. Phys. D, 462:134130, 2024.
Citations (3)
View on Semantic Scholar

Summary

We haven't generated a summary for this paper yet.

Ai Sparkles Streamline Icon: https://streamlinehq.com Summarize Now