Skip to content
/ nep2rat Public

Nep2rat is a tool to rationally approximate nonlinear matrix-valued functions, mainly aimed to solve nonlinear eigenvalue problems

Notifications You must be signed in to change notification settings

Gmnp/nep2rat

Repository files navigation

Nep2rat

'Nep2rat' is a tool to rationally approximate nonlinear matrix-valued functions, mainly aimed to solve nonlinear eigenvalue problems.

Contents

This repository contains the function "nep2rat" which approximates the nonlinear functions, and all the experiments proposed in the referenced paper. Those experiments require the installation of the NLEVP library.

Basic usage

 [Am, Bm, Rm] = nep2rat(F, Z)

returns the rational approximation Rm of the function F on the target set Z, and the linearized pencil (Am, Bm). The set Z must be given as a vector of points, while F could either be a function_handle or a struct, with fields:

  • F.coeffs: the matrix coefficients of F;
  • F.fun: the scalar functions that define f;

This structure mirrors the output of the NLEVP library. The default behaviour changes with the input F: if F is a function_handle, then nep2rat uses the "surrogate AAA with cyclic Leja--Bagby refinement" algorithm, while if F is a struct, it uses the "weighted AAA" algorithm. If the split form of F is available, we suggest to use the struct form of F as the input.

[Am, Bm, Rm, info] = nep2rat(F, Z, opts) 

allows the user to specify many optional parameters, such as the precision of the approximation and the algorithms utilised. The output info is returns additional information, while opts is a structure with all the. An in-depth guide is contained in the help of nep2rat.

References

Güttel, S., Negri Porzio, G.M. and Tisseur, F., 2020. Robust rational approximations of nonlinear eigenvalue problems. MIMS Eprints 2020.24

About

Nep2rat is a tool to rationally approximate nonlinear matrix-valued functions, mainly aimed to solve nonlinear eigenvalue problems

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Languages