Continuously
differentiable radial basis functions (C∞-RBFs), while
being theoretically exponentially convergent are considered impractical
computationally because the coefficient matrices are full and can become very ill-conditioned. Similarly, the
Hilbert and Vandermonde have full matrices and become ill-conditioned. The
difference between a coefficient matrix generated by C∞-RBFs for partial differential or integral equations and Hilbert and Vandermonde
systems is that C∞-RBFs are very sensitive to small changes in the
adjustable parameters. These parameters affect the condition number and
solution accuracy. The error terrain has many local and global maxima and
minima. To find stable and accurate numerical solutions for full linear
equation systems, this study proposes a hybrid combination of block Gaussian
elimination (BGE) combined with arbitrary precision arithmetic (APA) to
minimize the accumulation of rounding errors. In the future, this algorithm can
execute faster using preconditioners and implemented on massively parallel
computers.
References
[1]
Hardy, R.L. (1971) Multiquadric Equations of Topography and other Irregular Surfaces. Journal of Geophysical Research, 76, 1905-1915. https://doi.org/10.1029/JB076i008p01905
Kansa, E.J. (1990) Multiquadrics—A Scattered Data Approximation Scheme with Applications to Computational Fluid-Dynamics. II. Solutions to Parabolic, Hyperbolic and Elliptic Partial Differential Equations. Computers & Mathematics with Applications, 19, 147-161. https://doi.org/10.1016/0898-1221(90)90271-K
[4]
Macon, N. and Spitzbart, A. (February 1958) Inverses of Vandermonde Matrices. The American Mathematical Monthly, 65, 95-100. https://doi.org/10.1080/00029890.1958.11989147
[5]
Emdadi, A., Kansa, E.J., Ali Libre, N., Rahimian, M. and Shekarchi, M. (2008) Stable PDE Solution Methods for Large Multiquadric Shape Parameters. Computer Modeling in Engineering & Sciences, 25, 23-42.
[6]
Benzi, M., Haws, N.J. and Umas, M. (2000) Pre-Conditioning Highly Indefinite and Nonsymmetric Matrices. SIAM Journal on Scientific Computing, 22, 133-153. https://doi.org/10.1137/S1064827599361308
[7]
Ling, L. and Kansa, E.J. (2005) A Least Squares Preconditioner for Radial Basis Functions Collocation Methods. Advances in Computational Mathematics, 23, 31-54.
[8]
MPlapack. https://github.com/nakatamaho/mplapack
[9]
Holoborodko. https://www.advanpix.com
[10]
Dammel, J.W., Highman, N.J. and Schreiber, R. (1992) Block LU Factorization, NASA-BR-97949, Journal of Numerical Linear Algebra and Applications.
[11]
Eldersveld, S.K. and Saunders, M.A. (1992) A Block-LU Update for Large-Scale Linear Programming. SIAM Journal on Scientific Computing, 13, 191-201. https://doi.org/10.1137/0613016
[12]
Song, X., Jian, L. and Yang, S. (2017) A Chunk Updating LS-SVMs Based on Block Gaussian Elimination Method. Applied Soft Computing, 51, 96-104. https://doi.org/10.1016/j.asoc.2016.12.004
[13]
Hilbert, D. (1894) Ein Beitrag zur Theorie des Legendre’schen Polynoms. Acta Mathematica, 1, 155-159. https://doi.org/10.1007/BF02418278
[14]
Marcus, M. (1992) Vandermonde Matrix, §2.6.2. In: A Survey of Matrix Theory and Matrix Inequalities, Dover, New York, 15-16.
[15]
Highman, N.J. (1999) Fast Solution of Vandermonde-Like Systems Involving Orthogonal Polynomials. IMA Journal of Numerical Analysis, 8, 473-486. https://doi.org/10.1093/imanum/8.4.473
[16]
Kansa, E.J. and Holoborodko, P. (2017) On the Ill-Conditioned Nature of C∞ RBF Strong Collocation. Engineering Analysis with Boundary Elements, 78, 26-36. https://doi.org/10.1016/j.enganabound.2017.02.006
[17]
Galperin, E.A., Kansa, E.J., Makroglou. A. and Nelson S.A. (1997) Mathematical Programming Methods in the Numerical Solution of Volterra Integral and Integto-Differential Equations with Weakly-Singular Kernel. Nonlinear Analysis: Theory, Methods & Applications, 30, 1505-1513. https://doi.org/10.1016/S0362-546X(96)00340-9
[18]
Galperin, E.A. and Kansa, E.J. (2002) Application of Global Optimization and Radial Basis Functions to the Numerical Solution of Weakly Singular Volterra Integral Equations. Computers & Mathematics with Applications, 43, 439-456. https://doi.org/10.1016/S0898-1221(01)00300-5
[19]
Fedoseyev, A.I., Friedman, M.J. and Kansa, E.J. (2002) Improved Multiquadric Method for ell Partial Differential Equations via PDE Collocation on the Boundary. Computers & Mathematics with Applications, 43, 439-455. https://doi.org/10.1016/S0898-1221(01)00297-8
[20]
Kansa E.J. and Carlson, R.A. (1992) Improved Accuracy of Multiquadric Interpolation Using Variable Shape Parameters. Computers & Mathematics with Applications, 24, 99-120. https://doi.org/10.1016/0898-1221(92)90174-G
[21]
Wertz, J., Kansa, E.J. and Ling, L. (2006) The Role of the Multiquadric Shape Parameters in Solving Elliptic Partial Differential Equations. Computers & Mathematics with Applications, 51, 1335-1348. https://doi.org/10.1016/j.camwa.2006.04.009
[22]
Cervenka, M. and Skala, V. (2022) Conditionality Analysis of the Radial Basis Function Matrix. Computational Science and Its Applications—ICCSA 2020, Cagliari, 1-4 July 2020, 30-43. https://doi.org/10.1007/978-3-030-58802-1_3
[23]
Luh, L.T. (2016) The Mystery of the Shape Parameter IV. Engineering Analysis with Boundary Elements, 80, 103-109.
[24]
Luh, L.-T. (2019) The Choice of the Shape Parameter-A Friendly Approach. Engineering Analysis with Boundary Elements, 98, 103-109. https://doi.org/10.1016/j.enganabound.2018.10.011
[25]
Kansa, E.J. and Hon, Y.C. (2000) Circumventing the Ill-Conditioning Problem with Multiquadric Radial Basis Functions: Applications to Elliptic Partial Differential Equations. Computers & Mathematics with Applications, 39, 123-137. https://doi.org/10.1016/S0898-1221(00)00071-7
[26]
Hon, Y.C. and Schaback, R. (2002) On Unsymmetric Collocation by Radial Basis Functions. Applied Mathematics and Computation, 119, 177-188. https://doi.org/10.1016/S0096-3003(99)00255-6
