Constructing reductions for creative telescoping. The general differentially finite case (Q2666962)

Reduction-based algorithms for creative telescoping have been studied intensively in the past few years, and more and more general classes of functions have been investigated (rational, hyperexponential, algebraic, etc.) in the differential setting, and similarly in the shift case. The case of general holonomic functions in the differential case was missing, until in 2017/2018 two solutions to this problem were found independently: one is (a predecessor of) the present paper, the other one is by [\textit{A. Bostan} et al., in: Proceedings of the 43rd international symposium on symbolic and algebraic computation, ISSAC 2018. New York, NY: Association for Computing Machinery (ACM). 95--102 (2018; Zbl 1467.34011)]. To construct a reduction procedure for general holonomic functions, they are viewed as solutions of first-order differential systems (rather than higher-order scalar equations, however, it is well-known that both are equivalent). Then two types of reductions are defined: head reduction and tail reduction. The goal of the head reduction is to reduce the valuation of the given \(f\) at the point at infinity. For this purpose, special matrices, called ``head choppers'', are introduced. The tail reduction is supposed to reduce the valuation at a finite place \(\alpha\). It works analogous to the head reduction, by a change of variables. The combination of these two reductions gives the final algorithm. Some practical improvements (avoid expensive polynomial factoring by dynamic evaluation, and avoid computing explicitly with algebraic numbers) are discussed in the last section.