Abstract:We consider the asymptotic behavior of the expectation of the maximum for a special assignment process with constant or i.i.d. coefficients. We show how it depends on the coefficients' distribution.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is the asymptotic behavior of the maximum expectation in a specific random assignment process. Specifically, the author studies how the expectation of the maximum assignment value changes with the size of the matrix when the elements of the assignment matrix are constants or independent and identically distributed (i.i.d.) random variables, and how this change depends on the distribution of these random variables.
### Background of the Paper
In combinatorial optimization, the assignment problem (also known as the bipartite matching problem) is a classic problem. It involves a cost matrix \((w_{ij})_{1 \leq i, j \leq n}\), where \(w_{ij}\) represents the cost of assigning task \(i\) to executor \(j\). For each \(n\)-permutation \(\pi\), define a variable \(R_\pi=\sum_{i = 1}^n w_{i\pi(i)}\). The goal of the assignment problem is to find the permutation \(\pi\) that maximizes or minimizes \(R_\pi\).
If the cost matrix is random, the problem becomes a random assignment problem. It is usually assumed that \(w_{ij}\) are independent and identically distributed random variables, and the properties of the solution are studied in terms of how they depend on the common distribution of these random variables and the size of the matrix, especially the behavior when the matrix size tends to infinity.
### Research Content
This paper is particularly concerned with the expectation of the maximum value when the elements of the cost matrix \(w_{ij}=\log(1 + \gamma_{ij}h_{ij})\), where \(\gamma_{ij}\) and \(h_{ij}\) are independent non - negative random variables, and \(h_{ij}\) follows a standard exponential distribution. The author studies the asymptotic behavior of the expectation of the maximum value in this case and gives specific results for \(\gamma_{ij}\) under several different distributions:
1. **Constant \(\gamma\)**:
- When \(\gamma\) is a constant \(c\), the asymptotic behavior of the maximum expectation is \(E[\max_\pi R_\pi]\sim n\log\log n\).
2. **Exponential distribution \(\gamma\)**:
- When \(\gamma\) follows a standard exponential distribution, the asymptotic behavior of the maximum expectation is \(E[\max_\pi R_\pi]\sim 2n\log\log n\).
3. **Polynomial - tailed distribution \(\gamma\)**:
- When the density function of \(\gamma\) has a polynomial - tailed distribution \(q(y)\sim ay^{-\alpha}\) at infinity, the asymptotic behavior of the maximum expectation is \(E[\max_\pi R_\pi]\sim n(\log n)^{1/(\alpha - 1)}\).
4. **Uniform distribution \(\gamma\)**:
- When \(\gamma\) is uniformly distributed on \([0, 1]\), the asymptotic behavior of the maximum expectation is the same as in the case of a constant \(\gamma\), that is, \(E[\max_\pi R_\pi]\sim n\log\log n\).
### Main Conclusions
By analyzing \(\gamma\) under different distributions, the author obtains the asymptotic behavior of the maximum expectation and shows how these behaviors depend on the distribution of \(\gamma\). These results not only theoretically enrich the research on random assignment problems but also provide valuable references in practical applications, especially in fields such as wireless communication and resource allocation.