Abstract:Shifting the perspective from division points to division spacings leads not only to a substantially simpler proof of the conjecture but also to stronger results: limiting distributions are obtained for both partition points and spacings, with large-deviation error bounds. Maillard and Paquette's (2016) conjecture \cite{MaillardPaquette2016} is confirmed: the limiting empirical c.d.f. is independent of the splitting scheme.
The approach is naturally extended to arbitrary distributions of division points, with the limiting spacing distribution established. It is shown stationarity of the spacing distribution iff stationarity of the mean spacing iff uniform distribution of the partition points. This means that the results of Dean and Majumdar (2002) and Janson and Neininger (2008) concerning the limiting mean spacing apply directly to the limiting spacing distribution.
Discrete case (heavy-tailed distribution of division points) is studied and new results are presented, including the stage-wise progression of the fragmentation process. The results obtained apply to random trees.
From: Changqing Liu [view email]
[v1]
Sat, 16 Aug 2025 20:36:48 UTC (56 KB)
[v2]
Sat, 23 Aug 2025 00:56:07 UTC (59 KB)
[v3]
Tue, 1 Sep 2026 03:19:04 UTC (59 KB)
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。