Convergence of distributions on paths

Samy Abbes

DOI: https://doi.org/10.1007/978-3-031-43587-4_1

2024-10-02

Abstract:We study the convergence of distributions on finite paths of weighted digraphs, namely the family of Boltzmann distributions and the sequence of uniform distributions. Targeting applications to the convergence of distributions on paths, we revisit some known results from reducible nonnegative matrix theory and obtain new ones, with a systematic use of tools from analytic combinatorics. In several fields of mathematics, computer science and system theory, including concurreny theory, one frequently faces non strongly connected weighted digraphs encoding the elements of combinatorial structures of interest; this motivates our study.

Probability,Combinatorics

What problem does this paper attempt to address?

### What problems does this paper attempt to solve?

This paper mainly studies the convergence problem of path distributions on weighted digraphs. Specifically, the author focuses on two types of distributions on finite paths: Boltzmann distribution and uniform distribution, and explores their weak convergence behaviors under specific conditions. The following are the core problems of the paper:

1. **Convergence of path distributions in non - strongly - connected weighted digraphs**:
- For general, non - strongly - connected weighted digraphs, how to understand the asymptotic behavior of their path distributions? In particular, when the graph is not strongly connected, the classical Perron - Frobenius theory is no longer applicable, so new methods are needed to handle this situation.

2. **Weak convergence of Boltzmann distribution and uniform distribution**:
- In general weighted digraphs, do the Boltzmann distribution and the uniform distribution converge when the path length tends to infinity? If they converge, what type of measure do they converge to?
- In particular, under the periodicity assumption, does the uniform distribution also converge?

3. **Calculation of the residual matrix**:
- How to recursively calculate the residual matrix of a weighted digraph and use it to describe the asymptotic behavior of the path distribution?

4. **Study of umbrella digraphs**:
- Umbrella digraphs are a special type of digraphs, where the basic reachable class coincides with the final reachable class. What properties do these graphs have, and how can they be used to characterize the limit behavior of the path distribution?

5. **Existence of the complete cocycle measure**:
- Does a complete cocycle measure exist on umbrella digraphs, and what is the relationship between this measure and the limit of the path distribution?

Through the research of these problems, the author hopes to provide a new theoretical framework for non - strongly - connected weighted digraphs involved in fields such as mathematics, computer science, and system theory, in order to better understand and analyze the path distributions and their asymptotic behaviors in these graphs.

### Formula summary

- **Generating matrix \( H(z) \)**:
\[
H(z)=\sum_{k \geq 0} F^k z^k
\]
where \( F \) is the adjacency matrix of the weighted digraph.

- **Residual matrix \( \Theta \)**:
\[
\lim_{s \to \lambda}(1 - \lambda^{-1}s)^h H(s)=\Theta, \quad s \in(0, \lambda)
\]
where \( \lambda = \rho^{-1} \), and \( \rho \) is the spectral radius of the adjacency matrix.

- **Boltzmann distribution \( \theta_{x,s} \)**:
\[
\theta_{x,s}=\frac{1}{G_x(s)}\sum_{u \in O_x} w(u)s^{\ell(u)}\delta_{\{u\}}
\]
where \( G_x(s)=\sum_{u \in O_x} w(u)s^{\ell(u)} \).

- **Uniform distribution \( \mu_{x,k} \)**:
\[
\mu_{x,k}=\frac{1}{Z_x(k)}\sum_{u \in O_x(k)} w(u)\delta_{\{u\}}
\]
where \( Z_x(k)=\sum_{u \in O_x(k)} w(u) \).

- **Cocycle kernel \( q(x, y) \)**:
\[
q(x, y)=\rho^{-1} w(x, y)\Gamma(x, y)
\]
where \( \Gamma(x, y)=\frac{[\Theta \cdot 1]_y}{[\Theta \cdot 1]_x}