On the Approximation of Singular Functions by Series of Non-integer Powers

In this paper, we describe an algorithm for approximating functions of the form $f(x)=\int_{a}^{b} x^{\mu} \sigma(\mu) \, d \mu$ over $[0,1] \subset \mathbb{R}$, where $0<a<b<\infty$ and $\sigma(\mu)$ is some signed Radon measure over $[a,b]$ or some distribution supported on $[a,b]$. Given the desired accuracy $\epsilon$ and the values of $a$ and $b$, our method determines a priori a collection of non-integer powers $\{t_j\}_{j=1}^N$, so that the functions are approximated by series of the form $f(x)\approx \sum_{j=1}^N c_j x^{t_j}$, where the expansion coefficients can be found by solving a square, low-dimensional Vandermonde-like linear system using the collocation points $\{x_j\}_{j=1}^N$, also determined a priori by $\epsilon$ and the values of $a$ and $b$. We prove that our method has a small uniform approximation error which is proportional to $\epsilon$ multiplied by some small constants. We demonstrate the performance of our algorithm with several numerical experiments, and show that the number of singular powers and collocation points grows as $N=O(\log{\frac{1}{\epsilon}})$.