On variable and random shape Gaussian interpolations

Abstract This work focuses on the invertibility of non-constant shape Gaussian asymmetric interpolation matrix, which includes the cases of both variable and random shape parameters. We prove a sufficient condition for that these interpolation matrices are invertible almost surely for the choice of shape parameters. The proof is then extended to the case of anisotropic Gaussian kernels, which is subjected to independent componentwise scalings and rotations. As a corollary of our proof, we propose a parameter free random shape parameters strategy to completely eliminate the need of users’ inputs. By studying numerical accuracy in variable precision computations, we demonstrate that the asymmetric interpolation method is not a method with faster theoretical convergence. We show empirically in double precision, however, that these spatially varying strategies have the ability to outperform constant shape parameters in double precision computations. Various random distributions were numerically examined.

[1]  Scott A. Sarra,et al.  A random variable shape parameter strategy for radial basis function approximation methods , 2009 .

[2]  Leevan Ling,et al.  Fully adaptive kernel‐based methods , 2018 .

[3]  Song Xiang,et al.  Trigonometric variable shape parameter and exponent strategy for generalized multiquadric radial basis function approximation , 2012 .

[4]  David B. Dunson,et al.  Bayesian Data Analysis , 2010 .

[5]  Holger Wendland,et al.  Continuous and discrete least-squares approximation by radial basis functions on spheres , 2006, J. Approx. Theory.

[6]  Guang-Bin Huang,et al.  Trends in extreme learning machines: A review , 2015, Neural Networks.

[7]  Carl E. Rasmussen,et al.  Gaussian processes for machine learning , 2005, Adaptive computation and machine learning.

[8]  Fred J. Hickernell,et al.  Randomized Halton sequences , 2000 .

[9]  Benjamin Recht,et al.  Random Features for Large-Scale Kernel Machines , 2007, NIPS.

[10]  Liang Yan,et al.  Doubly stochastic radial basis function methods , 2018, J. Comput. Phys..

[11]  Nathan Halko,et al.  Finding Structure with Randomness: Probabilistic Algorithms for Constructing Approximate Matrix Decompositions , 2009, SIAM Rev..

[12]  Charles R. Johnson,et al.  Matrix analysis , 1985, Statistical Inference for Engineers and Data Scientists.

[14]  Bengt Fornberg,et al.  The Runge phenomenon and spatially variable shape parameters in RBF interpolation , 2007, Comput. Math. Appl..

[15]  Gregory E. Fasshauer,et al.  Kernel-based Approximation Methods using MATLAB , 2015, Interdisciplinary Mathematical Sciences.

[16]  Leevan Ling,et al.  A Fast Block-Greedy Algorithm for Quasi-optimal Meshless Trial Subspace Selection , 2016, SIAM J. Sci. Comput..

[17]  K. Życzkowski,et al.  Random unitary matrices , 1994 .

[18]  R. E. Carlson,et al.  Improved accuracy of multiquadric interpolation using variable shape parameters , 1992 .

[19]  Leevan Ling,et al.  Discrete least-squares radial basis functions approximations , 2019, Appl. Math. Comput..

[20]  Robert Schaback,et al.  On unsymmetric collocation by radial basis functions , 2001, Appl. Math. Comput..

[21]  Hamed Rabiei,et al.  On the new variable shape parameter strategies for radial basis functions , 2015 .

[22]  R. Schaback,et al.  Interpolation with variably scaled kernels , 2015 .