Guessing singular dependencies (Q2437325)

Given a complex function \(f(z)=\sum_k f_k z^k\), it is first shown how to compute its radius of convergence, its analytic continuation, the nearest singularity, and, possibly, to explore the Riemann surface beyond the singularity. When \(f\) is meromorphic, the methods can be simplified and alternatives are provided. These algorithms are needed in the main topic of the paper, which is the detection of analytic dependencies. If \(f\) (analytic in a closed disk except for a finite number of singularities) is algebraic at \(\sigma\), i.e., \(\sum_{k=0}^d P_k f^k=0\) with \(P_k\) analytic at \(\sigma\), an algorithm is provided to compute the \(\{P_k\}\). If \(f\) is Fuchsian, meaning that it satisfies a differential equation with polynomial coefficients \(L_k\) such that \(\sum_{k=0}^d L_k f^{(k)}=0\), again an algorithm to find the \(L_k\) is given. Alternatively, when analytic continuation is not an option, a numerical method is proposed for the following dependency problem. Given \(d\) formal power series \(f_k\) with radius of convergence at least \(r\), the problem is to find power series \(g_k\) and \(h\) such that \(\sum_{k=1}^d g_k f_k=h\). This is solved by truncating all power series after \(n\) terms, letting \(n\to\infty\) and solving the corresponding systen in a least squares sense. Numerical examples illustrate the methods.