The longest increasing subsequences of Luce permutations and permutons
Abstract
We investigate the longest increasing subsequences (LIS) of random permutations distributed under the Luce model with weights θ1(n) ≥ … ≥ θn(n). Our first main result provides a general formula for the growth rate of the LIS. For example, we get that the LIS of Luce permutations with Sukhatme weights θk(n) = n−k+1 grows like √n. We also give simple conditions for the LIS to be of order nβ, for any given β∈[1/2,1). Our second main result states that, if the weights are of the form θk(n) = f( (k−1)/n ) for a sufficiently regular function f, then the LIS divided by 2√n converges in probability to a constant K(f) described by a variational problem. Throughout, we make extensive use of coupling techniques; in particular, to prove our second main result, we argue that Luce random permutations behave similarly to random permutations sampled from the ``Luce permuton''.
Related
BibTeX
@misc{borga2026longest,
title = {{The longest increasing subsequences of Luce permutations and permutons}},
author = {Jacopo Borga and Victor Dubach},
howpublished = {HAL},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.30382},
eprint = {2610.30382},
archivePrefix = {arXiv},
url = {https://hal.science/hal-05837644v1},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - The longest increasing subsequences of Luce permutations and permutons
AU - Jacopo Borga
AU - Victor Dubach
PY - 2026
DA - 2026/10/21
PB - HAL
DO - 10.48550/arxiv.2610.30382
UR - https://hal.science/hal-05837644v1
AB - We investigate the longest increasing subsequences (LIS) of random permutations distributed under the Luce model with weights θ_1^(n) ≥ … ≥ θ_n^(n). Our first main result provides a general formula for the growth rate of the LIS. For example, we get that the LIS of Luce permutations with Sukhatme weights θ_k^(n) = n−k+1 grows like √n. We also give simple conditions for the LIS to be of order n^β, for any given β∈[1/2,1). Our second main result states that, if the weights are of the form θ_k^(n) = f( (k−1)/n ) for a sufficiently regular function f, then the LIS divided by 2√n converges in probability to a constant K(f) described by a variational problem. Throughout, we make extensive use of coupling techniques; in particular, to prove our second main result, we argue that Luce random permutations behave similarly to random permutations sampled from the ``Luce permuton''.
ER -