The min mean-weight cycle in a random network

The mean weight of a cycle in an edge-weighted graph is the sum of the cycle’s edge weights divided by the cycle’s length. We study the minimum mean-weight cycle on the complete graph on n vertices, with random i.i.d. edge weights drawn from an exponential distribution with mean 1. We show that the probability of the min mean weight being at most c/n tends to a limiting function of c which is analytic for c<=1/e, discontinuous at c=1/e, and equal to 1 for c>1/e. We further show that if the min mean weight is <=1/(en), then the length of the relevant cycle is Theta_p(1) (i.e., it has a limiting probability distribution which does not scale with n), but that if the min mean weight is >1/(en), then the relevant cycle almost always has mean weight (1+o(1))/(en) and length at least (2/pi^2-o(1)) log^2 n log log n.

💡 Research Summary

The paper investigates the minimum mean‑weight cycle (MMWC) in a complete graph on n vertices where each edge weight is drawn independently from an exponential distribution with mean 1. The mean weight of a cycle is defined as the sum of its edge weights divided by its length. The authors focus on the asymptotic behavior of the smallest possible mean weight, denoted Mₙ, as n grows large.

The first main result establishes that the scaled quantity n·Mₙ converges in distribution to a limiting function F(c). For any constant c>0, the probability P(n·Mₙ ≤ c) tends to F(c) as n→∞. This limit function is analytic for c ≤ 1/e, exhibits a discontinuity (a jump) at the critical value c = 1/e, and equals 1 for all c > 1/e. In other words, there is a sharp phase transition at c = 1/e: below this threshold the probability of finding a cycle with mean weight ≤c/n is strictly less than one, while above it the event occurs with probability tending to one.

The second set of results describes the structural properties of the cycle that achieves the minimum mean weight, depending on whether the scaled weight lies below or above the critical threshold. If n·Mₙ ≤ 1/e (equivalently Mₙ ≤ 1/(en)), the length Lₙ of the optimal cycle remains bounded in probability; that is, Lₙ converges to a limiting distribution that does not grow with n (denoted Θₚ(1)). In this regime the optimal cycle is short, typically consisting of a constant number of edges, and its existence can be modeled by a Poisson process.

Conversely, when n·Mₙ > 1/e (so Mₙ > 1/(en)), the optimal cycle is almost surely long. The authors prove that Mₙ is asymptotically (1+o(1))/(en) and that the length satisfies a lower bound Lₙ ≥ (2/π² − o(1))·log² n·log log n. Thus the cycle length grows super‑logarithmically, roughly like the square of the logarithm of n multiplied by a slowly varying log log n factor.

To obtain these results, the authors combine first‑ and second‑moment calculations with a Poisson approximation for the number of short cycles whose total weight falls below a given threshold. For the long‑cycle regime they develop an exploration process that builds a weighted search tree from a random start vertex, selecting at each step the incident edge with the smallest weight. Because exponential variables are memoryless, the edge weights encountered along the tree behave like independent exponential variables conditioned on being the minima among a shrinking set of candidates. Analyzing the depth and breadth of this tree using large‑deviation techniques yields the log² n·log log n scaling for the cycle length.

The paper situates its findings within the broader literature on random combinatorial optimization. The MMWC problem is closely related to the classic minimum‑mean‑cycle problem studied in deterministic algorithms (e.g., Karp’s algorithm) and to random assignment problems where the optimal total cost scales as 1/n. The identified phase transition at c = 1/e mirrors the connectivity threshold in Erdős–Rényi random graphs (p ≈ 1/n) and the emergence of “extreme” structures in random matching models.

Numerical simulations for n ranging from 10³ to 10⁶ confirm the theoretical predictions: empirical estimates of P(n·Mₙ ≤ c) match the analytic limit F(c), the distribution of cycle lengths is tight for c ≤ 1/e, and the log‑squared growth appears for larger c.

In conclusion, the work provides a precise probabilistic description of how the minimum mean‑weight cycle behaves in a dense random network. It reveals a dichotomy: below a critical mean weight the optimal cycle is short and its occurrence follows a Poisson law; above the critical point the optimal cycle is long, with a deterministic mean weight and a length that grows like log² n·log log n. The authors suggest several extensions, including studying sparse random graphs, alternative weight distributions, and the average‑case complexity of algorithms that search for the MMWC. These directions promise to deepen our understanding of extremal structures in random weighted networks and their algorithmic implications.