Differential rational normal forms and a reduction algorithm for hyperexponential func

We describe differential rational normal forms of a rational function and their properties. Based on these normal forms, we present an algorithm which, given a hyperexponential function T(x), constructs two hyperexponential functions T;<sub>1;</sub>(x) and T;<sub>2;</sub>(x) such that T(x) = T;<sub>1;</sub><sup>'</sup>(x) + T;<sub>2;</sub>(x) and T;<sub>2;</sub>(x) is minimal in some sense. The algorithm can be used to accelerate the differential Gosper's algorithm and to compute right factors of the telescopers.