Table of Contents
Fetching ...

About the Moments of the Generalized Ulam Problem

Samen Hossein, Shannon Starr

TL;DR

This work addresses the generalized Ulam problem, focusing on the moments and large-deviation behavior of $Z_{n,k}$, the number of increasing subsequences of length $k$ in a random permutation. It derives an exact second-moment decomposition $\mathbb{E}[Z_{n,k}Z_{n,\ell}] = \sum_{j=0}^{k\wedge\ell} \mathbb{E}[Z_{n,k+\ell-j}] \mathcal{A}(k-j,\ell-j,j)$ with a combinatorial factor $\mathcal{A}$ whose multivariate generating function is $\sum \mathcal{A}(k,\ell,j) x^k y^\ell z^j = 1/\sqrt{1-2(x+y)+(x-y)^2}-z$, enabling a multivariate saddle-point analysis in the critical regime $k,\ell \sim \mathcal{O}(\sqrt{n})$. The paper also introduces an exactly solvable model for sums of i.i.d. exponentials, develops a replica-symmetric ansatz and a replica-to-zero trick to access leading-order behavior of higher moments, and reveals limitations of replica-symmetric predictions via a Poisson-point-process description near the minimum. Extending to the third moment, it employs diagonal generating-function methods to obtain a generating function for $r=3$ that reduces to an elliptic integral, and discusses interpretations in terms of random-walk intersections in $d=2$ and $d=3$. Overall, the results connect combinatorial moment calculations with analytic and probabilistic physics-inspired techniques, and lay groundwork for future rigorous validation and extensions to higher moments.

Abstract

Given $π\in S_n$, let $Z_{n,k}(π)=\sum_{1\leq i_1<\dots<i_k\leq n} \mathbf{1}(\{ π_{i_1}<\dots<π_{i_k}\}$ denote the number of increasing subsequences of length $k$. Consider the "generalized Ulam problem," studying the distribution of $Z_{n,k}$ for general $k$ and $n$. For the 2nd moment, Ross Pinsky initiated a combinatorial study by considering a pair of subsequences $i^{(r)}_1<\dots<i^{(r)}_k$ for $r \in \{1,2\}$, and conditioning on the size of the intersection $j = |\{i_1^{(1)},\dots,i^{(1)}_k\} \cap \{i^{(2)}_1,\dots,i^{(2)}_k\}|$. We obtain the exact large deviation rate function for $\mathbf{E}[Z_{n,k} Z_{n,\ell}]$ in the asymptotic regime $k\sim κn^{1/2}$, $\ell \sim λn^{1/2}$ as $n \to \infty$, for $κ,λ\in (0,\infty)$. This uses multivariate generating function techniques, as found in the textbook of Pemantle and Wilson. The requisite generating function enumerates pairs of up-right paths in $d=2$, which both end at $(k,\ell)$ with a given number of intersections. We also evaluate the analogous generating function for pairs of $(+\boldsymbol{i},+\boldsymbol{j},+\boldsymbol{k})$ paths in $d=3$, which both end at $(k,\ell,m)$, which has some utility in calculating the 3rd moment. Finally, we consider a simpler problem involving partitions instead of permutations, where all moments are calculable and the replica symmetric ansatz can be stated if not proved.

About the Moments of the Generalized Ulam Problem

TL;DR

This work addresses the generalized Ulam problem, focusing on the moments and large-deviation behavior of , the number of increasing subsequences of length in a random permutation. It derives an exact second-moment decomposition with a combinatorial factor whose multivariate generating function is , enabling a multivariate saddle-point analysis in the critical regime . The paper also introduces an exactly solvable model for sums of i.i.d. exponentials, develops a replica-symmetric ansatz and a replica-to-zero trick to access leading-order behavior of higher moments, and reveals limitations of replica-symmetric predictions via a Poisson-point-process description near the minimum. Extending to the third moment, it employs diagonal generating-function methods to obtain a generating function for that reduces to an elliptic integral, and discusses interpretations in terms of random-walk intersections in and . Overall, the results connect combinatorial moment calculations with analytic and probabilistic physics-inspired techniques, and lay groundwork for future rigorous validation and extensions to higher moments.

Abstract

Given , let denote the number of increasing subsequences of length . Consider the "generalized Ulam problem," studying the distribution of for general and . For the 2nd moment, Ross Pinsky initiated a combinatorial study by considering a pair of subsequences for , and conditioning on the size of the intersection . We obtain the exact large deviation rate function for in the asymptotic regime , as , for . This uses multivariate generating function techniques, as found in the textbook of Pemantle and Wilson. The requisite generating function enumerates pairs of up-right paths in , which both end at with a given number of intersections. We also evaluate the analogous generating function for pairs of paths in , which both end at , which has some utility in calculating the 3rd moment. Finally, we consider a simpler problem involving partitions instead of permutations, where all moments are calculable and the replica symmetric ansatz can be stated if not proved.
Paper Structure (36 sections, 5 theorems, 217 equations, 1 figure)

This paper contains 36 sections, 5 theorems, 217 equations, 1 figure.

Key Result

Proposition 2.1

For any $n \in \mathbb{N} = \{1,2,\dots\}$ and any $k \in [n]$, it is known $\mathbf{E}[Z_{n,k}] = \binom{n}{k} / k!$. Now, for any $k,\ell \in [n]$, where

Figures (1)

  • Figure 1: The diagonal transformation that is part of Pinsky's combinatorial theorem. The choice of the next step being among $\{(0,1),(1,0),(-1,0),(0,-1)\}$ is equivalent to a diagonal version of the encoding being in the set $\{(+1,+1),(+1,-1),(-1,-1),(-1,+1)\}$ as shown.

Theorems & Definitions (5)

  • Proposition 2.1
  • Corollary 2.2
  • Corollary 2.3
  • Theorem 2.4
  • Lemma 2.5