In, 1751, Euler was studying the number of ways in which a given convex polygon could be decomposed into triangles by diagonal lines. (Flajolet & Sedgewick, p.20)
He realized that the progression of numbers in the solution (1, 2, 5, 14, 42, 132,…) was directly related to the coefficients of the series expansion of the polynomial fraction (1−2a−√(1−4a)) / (2aa), that is: 1+2a +5a^2 + 14a^3 + 42a^4 + 132a^5 + …
Given any constructable combinatorial structure, one can use a set of operators to find a generating function and then approach the problem analytically.
This seems like an interesting package. Unfortunately, I lack the expertise to help. Nonetheless, I wonder whether you can find any inspiration from Symbolics.jl.
Version 2.0 is out with major performance enhancements
I needed to run some simulations with large samples and SymPy.jl was unfeasible.
I changed some of the methods (SymPy :Sym expressions for Julia native functions and SymPy.series for TaylorSeries.series_expand).
Run times were 10 to 10^5 faster (no kidding) for some methods . I am quite happy with Julia at the moment.
If this package seems interesting, stay in tune.
I’ll release SymbolicInference.jl soon. It uses AnalyticComb.jl to make inference on real problems (e.g. chaotic time-series). Nils (NilsToAn · GitHub) and Norbert Marwan (pucicu (Norbert Marwan) · GitHub) from Potsdam started contributing on a paper about this.
For example, knowing the generating function of a sequence, and therefore all its singular varieties, can your code provide assymptotics on the sequence itself ? I think the book provide results on both first and second order if I am right. Is that implemented or is that in the “second part” ?