A novel algorithm for best rational approximation with applications to fractional diffusion

Dr. Clemens Hofreither

Oct. 13, 2020, 1:30 p.m. S2 054

It has recently been understood that there is a deep connection between rational approximation of functions of the type $x \mapsto x^\alpha$ and numerical methods for elliptic fractional diffusion problems. However, in particular the computation of best rational approximations in the maximum norm of these functions has posed a significant obstacle in the practical realization of numerical methods exploiting this link.

To overcome these difficulties, a novel algorithm for computing best uniform rational approximations to real scalar functions in the setting of zero defect is proposed in this talk. The method, dubbed BRASIL (best rational approximation by successive interval length adjustment), is based on the observation that the best rational approximation $r$ to a function $f$ must interpolate $f$ at a certain number of interpolation nodes $(x_j)$. Furthermore, the sequence of local maximum errors per interval $(x_{j-1},x_j)$ must equioscillate. The proposed algorithm iteratively rescales the lengths of the intervals with the goal of equilibrating the local errors. The required rational interpolants are computed in a stable way using the barycentric rational formula.

The new algorithm exhibits excellent numerical stability and computes best rational approximations of high degree to many functions in a few seconds, using only standard IEEE double-precision arithmetic. In some examples, we demonstrate that it converges quickly in some situations where the current state-of-the-art method, the $\texttt{minimax}$ function from the $\texttt{Chebfun}$ package, fails.