DOI QR코드

DOI QR Code

LEARNING-BASED NUMERICAL METHODS FOR THE NONLINEAR MATRIX EQUATION X2 = A.

  • Hyun-Min Kim (School of Liberal Studies, Department of Mathematics, Industrial Mathematics Center, and Humanoid olfactory display Innovation research Center, Pusan National University)
  • 투고 : 2026.05.06
  • 심사 : 2026.05.21
  • 발행 : 2026.05.31

초록

We study the problem of approximating the principal square root A1/2 of a nonsingular matrix A ∈ ℂn×n with no eigenvalues on ℝ≤0. Classical iterative methods-Newton, Denman-Beavers, Halley, productform, Padé, and the Schur method of Björck-Hammarling-each require O(n3) work per step and may converge slowly near the branch cut. We cast the map F : A 7→ A1/2 as an empirical risk minimization problem over deep ReLU networks, establish a Rademacher-complexity generalization bound, and show that the principal square-root map is qualitatively approximable by ReLU networks on the compact spectral band An,α,β. Two hybrid algorithms are proposed in which a trained network warm-starts Newton or Denman-Beavers refinement; their iteration savings follow directly from the quadratic convergence of Newton's method without new convergence proofs.

키워드

과제정보

This work was supported by a 2-Year Research Grant of Pusan National University.

참고문헌

  1. A. R. Barron, Universal approximation bounds for superpositions of a sigmoidal function, IEEE Trans. Inform. Theory 39(3) (1993), 930-945. https://doi.org/10.1109/TIT.18
  2. P. L. Bartlett, D. J. Foster and M. J. Telgarsky, Spectrally-normalized margin bounds for neural networks, in Advances in Neural Information Processing Systems (NeurIPS), Vol. 30, 2017.
  3. Å. Björck and S. Hammarling, A Schur method for the square root of a matrix, Linear Algebra Appl. 52-53 (1983), 127-140. https://doi.org/10.1016/0024-3795(83)90010-1
  4. C. S. Chacha and H.-M. Kim, Elementwise minimal nonnegative solutions for a class of nonlinear matrix equations, East Asian J. Appl. Math. 9(4) (2019), 665-682. https://doi.org/10.4208/eajam
  5. S. H. Cheng, N. J. Higham, C. S. Kenney and A. J. Laub, Approximating the logarithm of a matrix to specified accuracy, SIAM J. Matrix Anal. Appl. 22(4) (2001), 1112-1125. https://doi.org/10.1137/S0895479899364015
  6. G. Cybenko, Approximation by superpositions of a sigmoidal function, Math. Control Signals Systems 2 (1989), 303-314. https://doi.org/10.1007/BF02551274
  7. E. D. Denman and A. N. Beavers, The matrix sign function and computations in systems, Appl. Math. Comput. 2(1) (1976), 63-94. https://doi.org/10.1016/0096-3003(76)90020-5
  8. N. J. Higham, Functions of Matrices: Theory and Computation, SIAM, Philadelphia, PA, 2008.
  9. N. J. Higham and H.-M. Kim, Numerical analysis of a quadratic matrix equation, IMA J. Numer. Anal. 20(4) (2000), 499-519. https://doi.org/10.1093/imanum/20.4.499
  10. N. J. Higham and H.-M. Kim, Solving a quadratic matrix equation by Newton's method with exact line searches, SIAM J. Matrix Anal. Appl. 23(2) (2001), 303-316. https://doi.org/10.1137/S0895479899350976
  11. B. Iannazzo, A note on computing the matrix square root, Calcolo 40(4) (2003), 273-283. https://doi.org/10.1007/s10092-003-0079-9
  12. Y.-J. Kim, H.-M. Kim and J. Meng, Newton-Schulz method for solving a certain nonlinear matrix equation, J. Korean Math. Soc. 55 (2018), 1529-1540.
  13. T. Kim, H.-M. Kim and S.-H. Seo, On Newton's method for solving a system of nonlinear matrix equations, East Asian Math. J. 35(3) (2019), 341-349.
  14. N. Kovachki, Z. Li, B. Liu, K. Azizzadenesheli, K. Bhattacharya, A. Stuart and A. Anandkumar, Neural operator: Learning maps between function spaces with applications to PDEs, J. Mach. Learn. Res. 24(89) (2023), 1-97.
  15. H. Lee, H.-M. Kim and J. Meng, On the nonlinear matrix equation X + A*X-qA = Q, J. Comput. Appl. Math. 373 (2020).
  16. L. Lu, P. Jin, G. Pang, Z. Zhang and G. E. Karniadakis, Learning nonlinear operators via DeepONet based on the universal approximation theorem of operators, Nature Machine Intelligence 3 (2021), 218-229. https://doi.org/10.1038/s42256-021-00302-5
  17. J. Meng and H.-M. Kim, The positive definite solution of a certain nonlinear matrix equation, Numer. Funct. Anal. Optim. 39 (2018), 398-412. https://doi.org/10.1080/01630563.2017.1369108
  18. J. Meng and H.-M. Kim, Normwise, mixed and componentwise condition numbers of a matrix equation, Oper. Matrices 13(1) (2019), 19-34. https://doi.org/10.7153/oam-2019-13-02
  19. J. Meng, H.-M. Kim and S.-H. Seo, Condition numbers and backward error of a matrix polynomial equation arising in stochastic models, J. Sci. Comput. 76 (2018), 759-776. https://doi.org/10.1007/s10915-018-0641-x
  20. J. Meng, H. Kim and H. J. Chen, A further study on a nonlinear matrix equation, Japan J. Indust. Appl. Math. 37(3) (2020), 831-849. https://doi.org/10.1007/s13160-020-00421-3
  21. J. Meng, H. Lee and H.-M. Kim, On the extreme solutions of a class of nonlinear matrix equation, J. Nonlinear Convex Anal. 21(1) (2020), 77-87.
  22. C. Moler and C. Van Loan, Nineteen dubious ways to compute the exponential of a matrix, SIAM Rev. 20(4) (1978), 801-836. https://doi.org/10.1137/1020098
  23. M. Mohri, A. Rostamizadeh and A. Talwalkar, Foundations of Machine Learning, 2nd ed., MIT Press, Cambridge, MA, 2018.
  24. S. M. R. S. Naqvi, J. Meng and H.-M. Kim, Iteration with stepsize parameter and condition numbers for a nonlinear matrix equation, Electron. J. Linear Algebra 34 (2018), 217-230. https://doi.org/10.13001/1081-3810.3449
  25. J.-H. Seo and H.-M. Kim, Convergence of pure and relaxed Newton methods for solving a matrix polynomial equation arising in stochastic models, Linear Algebra Appl. 440(1) (2014), 34-49. https://doi.org/10.1016/j.laa.2013.10.043
  26. S.-H. Seo, H.-M. Kim and J. Meng, Convergence of a modified Newton method for a matrix polynomial equation arising in stochastic problems, Electron. J. Linear Algebra 34 (2018), 500-513. https://doi.org/10.13001/1081-3810.3762