Sinc numerical methods

In numerical analysis and applied mathematics, sinc numerical methods are numerical techniques[1] for finding approximate solutions of partial differential equations and integral equations based on the translates of sinc function and Cardinal function C(f,h) which is an expansion of f defined by

C(f,h)(x)=\sum _{k=-\infty }^{\infty }f(kh)\,{\textrm {sinc}}\left({\dfrac {x}{h}}-k\right)

where the step size h>0 and where the sinc function is defined by

{\textrm {sinc}}(x)={\frac {\sin(\pi x)}{\pi x}}

Sinc approximation methods excel for problems whose solutions may have singularities, or infinite domains, or boundary layers.

The truncated Sinc expansion of f is defined by the following series:

C_{M,N}(f,h)(x)=\displaystyle \sum _{k=-M}^{N}f(kh)\,{\textrm {sinc}}\left({\dfrac {x}{h}}-k\right)

.

Sinc numerical methods cover

function approximation,
approximation of derivatives,
approximate definite and indefinite integration,
approximate solution of initial and boundary value ordinary differential equation (ODE) problems,
approximation and inversion of Fourier and Laplace transforms,
approximation of Hilbert transforms,
approximation of definite and indefinite convolution,
approximate solution of partial differential equations,
approximate solution of integral equations,
construction of conformal maps.

Indeed, Sinc are ubiquitous for approximating every operation of calculus

In the standard setup of the sinc numerical methods, the errors (in big O notation) are known to be $O\left(e^{-c{\sqrt {n}}}\right)$ with some c>0, where n is the number of nodes or bases used in the methods. However, Sugihara[2] has recently found that the errors in the Sinc numerical methods based on double exponential transformation are $O\left(e^{-{\frac {kn}{\ln n}}}\right)$ with some k>0, in a setup that is also meaningful both theoretically and practically and are found to be best possible in a certain mathematical sense.

Reading

Stenger, Frank (2011). Handbook of Sinc Numerical Methods. Boca Raton, Florida: CRC Press. ISBN 9781439821596.
Lund, John; Bowers, Kenneth (1992). Sinc Methods for Quadrature and Differential Equations. Philadelphia: Society for Industrial and Applied Mathematics (SIAM). ISBN 9780898712988.

gollark: There's "feedforward CLIP VQGAN" or something which is *much* faster but worse.

gollark: Anyway, excluding the tensor cores and VRAM, standard colab's T4s are worse than a 1080 Ti. The K80 is worse than most recent things and you often get those now.

gollark: If you use it heavily you may as well just buy a P100 on eBay, they're "just" £600 or so now.

gollark: It's worse now! There's a Pro+ for 5 times as much and Pro people mostly get T4s.

gollark: 700-series, not 1000.

References

Stenger, F. (2000). "Summary of sinc numerical methods". Journal of Computational and Applied Mathematics. 121: 379–420. doi:10.1016/S0377-0427(00)00348-4.
Sugihara, M.; Matsuo, T. (2004). "Recent developments of the Sinc numerical methods". Journal of Computational and Applied Mathematics. 164-165: 673. doi:10.1016/j.cam.2003.09.016.

This article is issued from Wikipedia. The text is licensed under Creative Commons - Attribution - Sharealike. Additional terms may apply for the media files.

[1] Stenger, F. (2000). "Summary of sinc numerical methods". Journal of Computational and Applied Mathematics. 121: 379–420. doi:10.1016/S0377-0427(00)00348-4.

[2] Sugihara, M.; Matsuo, T. (2004). "Recent developments of the Sinc numerical methods". Journal of Computational and Applied Mathematics. 164-165: 673. doi:10.1016/j.cam.2003.09.016.