Generalized Hofstadter functions $G, H$ and beyond: numeration systems and discrepancy
Pierre Letouzey (IRIF, PICUBE)

TL;DR
This paper explores generalized Hofstadter functions related to Fibonacci-like sequences, their digital expansions, discrepancy properties, and solves longstanding conjectures, with formal verification in Coq/Rocq.
Contribution
It introduces a family of generalized Hofstadter functions, analyzes their discrepancy, and proves conjectures about their approximation by linear functions, all formalized in proof assistants.
Findings
Discrepancy is finite for k ≤ 4.
F_3 and F_4 are close to their linear parts, confirming conjectures.
F_k coincides with linear parts only for k ≤ 2.
Abstract
Hofstadter's function is recursively defined via and then . Following Hofstadter, a family of similar functions is obtained by varying the number of nested recursive calls in this equation. We study here some Fibonacci-like sequences that are deeply connected with these functions . In particular, the Zeckendorf theorem can be adapted to provide digital expansions via sums of terms of these sequences. On these digital expansions, the functions are acting as right shifts of the digits. These Fibonacci-like sequences can be expressed in terms of zeros of the polynomial . Considering now the discrepancy of each function , i.e., the maximal distance between and its linear equivalent, we retrieve the fact that this discrepancy is finite exactly when . Thanks to that, we solve two twenty-year-old OEIS…
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
