Min Max Average Cycle Weight Optimization

• Min Max Average Cycle Weight is a cycle-based optimization objective that seeks to minimize the maximum average weight across cycles in weighted directed graphs.
• It unifies methods from minimum mean cycle problems, rooted min-max cycle covers, and fair allocation via envy graphs, each with specialized algorithms and analyses.
• Research in this area addresses stochastic threshold phenomena in random graphs alongside exact and approximate algorithmic solutions, highlighting practical tradeoffs.

“Min Max Average Cycle Weight” denotes a family of cycle-based optimization objectives centered on extremal cycle weights. In classical graph optimization, the closest standard notion is the minimum mean-weight cycle, which asks for a directed cycle minimizing total weight divided by length, $\mu(C)=w(C)/|C|$ (Chatterjee et al., 2013). In more recent fair-allocation work, the same phrase is used more literally for the problem of choosing an allocation that minimizes the maximum average weight of directed cycles in an induced envy graph, that is, $\min_A MACW(G_A)$ with $MACW(G)=\max_C \overline w(C)$ (Elmalem et al., 27 Jul 2025). Related literature also studies min-max cycle covers, where one minimizes the maximum total weight of any cycle in a cover rather than a cycle mean (Lin et al., 2020).

1. Core definitions and objective families

For a weighted directed graph $G=(V,E,w)$, a cycle $C$ has total weight

$w(C)=\sum_{e\in C} w(e),$

length $|C|$, and mean or average weight

$\mu(C)=\frac{w(C)}{|C|}.$

The minimum cycle mean or minimum average cycle weight problem is

$\mu^*=\min_C \mu(C).$

A dual maximum version is

$\mu_{\max}=\max_C \mu(C),$

and exact minimum and maximum versions are equivalent under weight negation: if $\min_A MACW(G_A)$0, then

$\min_A MACW(G_A)$1

(Chatterjee et al., 2013).

A distinct but related line studies families of cycles rather than a single cycle. In rooted cycle-cover problems, a cycle cover $\min_A MACW(G_A)$2 has maximum cycle weight

$\min_A MACW(G_A)$3

total cycle weight

$\min_A MACW(G_A)$4

and average cycle weight

$\min_A MACW(G_A)$5

where $\min_A MACW(G_A)$6 is the number of cycles. The rooted min-max cycle cover problem optimizes $\min_A MACW(G_A)$7, not $\min_A MACW(G_A)$8, although $\min_A MACW(G_A)$9 for every cover (Lin et al., 2020).

A third formulation appears in envy-graph optimization. Given an allocation $MACW(G)=\max_C \overline w(C)$0, the envy of agent $MACW(G)=\max_C \overline w(C)$1 at agent $MACW(G)=\max_C \overline w(C)$2 is

$MACW(G)=\max_C \overline w(C)$3

which induces a complete directed envy graph $MACW(G)=\max_C \overline w(C)$4. For any directed cycle $MACW(G)=\max_C \overline w(C)$5, its average weight is

$MACW(G)=\max_C \overline w(C)$6

and the graph functional is

$MACW(G)=\max_C \overline w(C)$7

The optimization problem is then

$MACW(G)=\max_C \overline w(C)$8

or, with preexisting conditions, $MACW(G)=\max_C \overline w(C)$9 (Elmalem et al., 27 Jul 2025).

Objective	Optimization form	Representative setting
Minimum mean-weight cycle	$G=(V,E,w)$0	weighted digraphs, random complete graphs
Rooted min-max cycle cover	$G=(V,E,w)$1	multi-robot routing
Min Max Average Cycle Weight	$G=(V,E,w)$2	envy graphs in fair allocation

2. Threshold behavior in random complete graphs

A particularly detailed probabilistic theory is available for the minimum mean-weight cycle on the complete graph or complete digraph with i.i.d. exponential edge weights of mean $G=(V,E,w)$3. In this stochastic mean-field distance model, the natural scale of the minimum average cycle weight is $G=(V,E,w)$4, and Mathieu and Wilson showed a sharp threshold at $G=(V,E,w)$5 (Mathieu et al., 2012).

Let $G=(V,E,w)$6 denote the minimum mean cycle weight. For the directed complete graph, the limiting probability of a $G=(V,E,w)$7-light cycle satisfies

$G=(V,E,w)$8

where the tree function is

$G=(V,E,w)$9

For undirected complete graphs, the corresponding limit is

$C$0

for $C$1, and $C$2 for $C$3. The limiting distribution is analytic for $C$4, discontinuous at $C$5, and equal to $C$6 for $C$7 (Mathieu et al., 2012).

The same work identified a two-regime picture for the minimizing cycle. If the minimum mean weight is at most $C$8, the minimizing cycle has length $C$9 and fixed-length probabilities $w(C)=\sum_{e\in C} w(e),$0 converge to explicit integrals. If the minimum mean weight is larger than $w(C)=\sum_{e\in C} w(e),$1, then conditional on that event,

$w(C)=\sum_{e\in C} w(e),$2

with probability $w(C)=\sum_{e\in C} w(e),$3 (Mathieu et al., 2012).

Ding, Sun, and Wilson completed the supercritical analysis. Writing $w(C)=\sum_{e\in C} w(e),$4 for the minimum mean weight and $w(C)=\sum_{e\in C} w(e),$5 for the length of the minimizing cycle, and conditioning on $w(C)=\sum_{e\in C} w(e),$6, they proved

$w(C)=\sum_{e\in C} w(e),$7

with high probability, equivalently

$w(C)=\sum_{e\in C} w(e),$8

and

$w(C)=\sum_{e\in C} w(e),$9

Thus the supercritical minimum average cycle weight sits just above $|C|$0, with an explicit $|C|$1 correction, and the minimizing cycle is polylogarithmically long rather than constant-sized (Ding et al., 2015).

The mechanism behind the threshold is a first-moment balance for $|C|$2-light $|C|$3-cycles together with confinement of an associated exp-minus-one random walk. In the refined supercritical analysis, the principal eigenvalue

$|C|$4

of the exp-minus-one walk killed outside $|C|$5 produces the same $|C|$6 correction that appears in $|C|$7 (Ding et al., 2015).

3. Exact and approximate algorithms for minimum cycle mean

The algorithmic minimum cycle mean problem is standard for directed graphs with integer weights. One formulation uses

$|C|$8

and seeks

$|C|$9

A dynamic-programming characterization is based on path weights of exact length $\mu(C)=\frac{w(C)}{|C|}.$0: if $\mu(C)=\frac{w(C)}{|C|}.$1 is the minimum weight of a length-$\mu(C)=\frac{w(C)}{|C|}.$2 path starting at $\mu(C)=\frac{w(C)}{|C|}.$3, then

$\mu(C)=\frac{w(C)}{|C|}.$4

where $\mu(C)=\frac{w(C)}{|C|}.$5 is the minimum mean weight of a cycle reachable from $\mu(C)=\frac{w(C)}{|C|}.$6 (Chatterjee et al., 2013).

Classical exact algorithms include Karp’s $\mu(C)=\frac{w(C)}{|C|}.$7 method, and the literature also includes parametric shortest-path methods such as Karp–Orlin, improved by Young–Tarjan–Orlin (Ding et al., 2015). In the matrix-algebraic direction, the minimum cycle mean problem reduces to min-plus matrix multiplication. One result shows that exact minimum cycle mean is reducible in $\mu(C)=\frac{w(C)}{|C|}.$8 time to a logarithmic number of min-plus matrix multiplications of $\mu(C)=\frac{w(C)}{|C|}.$9 matrices, and that for nonnegative weights there is a $\mu^*=\min_C \mu(C).$0-approximation algorithm with runtime

$\mu^*=\min_C \mu(C).$1

where $\mu^*=\min_C \mu(C).$2 is the maximum weight and $\mu^*=\min_C \mu(C).$3 is the time for ordinary matrix multiplication (Chatterjee et al., 2013).

A later development revisited minimum mean cycle on low-diameter graphs. Altschuler and Parrilo gave an approximation algorithm based on a linear programming relaxation, entropic regularization, and reduction to matrix balancing. For graphs with polylogarithmic diameter, the runtime is near-linear in the number of edges, and for complete graphs it is the first algorithm whose dependence on $\mu^*=\min_C \mu(C).$4 is $\mu^*=\min_C \mu(C).$5. The method uses only $\mu^*=\min_C \mu(C).$6 memory beyond reading the input, and returns a cycle $\mu^*=\min_C \mu(C).$7 with

$\mu^*=\min_C \mu(C).$8

(Altschuler et al., 2020).

The algorithmic distinction between minimum total cycle weight and minimum mean cycle weight remains important. Strongly polynomial $\mu^*=\min_C \mu(C).$9-approximation schemes are known for minimum-weight cycle via approximate APSP and min-plus methods, but those schemes target total cycle weight, not the normalized mean objective (Bringmann et al., 2019). This suggests that mean-cycle and sum-cycle problems share algebraic infrastructure but are not interchangeable.

4. Min-max cycle covers and worst-cycle objectives

In rooted multi-robot routing, the central optimization target is often not a cycle mean but a worst-route criterion. One such formulation is the rooted min-max cycle cover problem on a complete undirected graph

$\mu_{\max}=\max_C \mu(C),$0

with nonnegative metric edge weights $\mu_{\max}=\max_C \mu(C),$1. The vertex set contains depots $\mu_{\max}=\max_C \mu(C),$2 and inspection sites $\mu_{\max}=\max_C \mu(C),$3. A rooted cycle cover $\mu_{\max}=\max_C \mu(C),$4 must satisfy: $\mu_{\max}=\max_C \mu(C),$5, the cycles are edge-disjoint, each cycle contains exactly one depot, and the union of cycle vertices is all of $\mu_{\max}=\max_C \mu(C),$6 (Lin et al., 2020).

For a cycle $\mu_{\max}=\max_C \mu(C),$7,

$\mu_{\max}=\max_C \mu(C),$8

and the optimization problem is

$\mu_{\max}=\max_C \mu(C),$9

In travel-time applications this minimizes makespan; in energy applications it minimizes the maximum energy used by any robot. The paper explicitly distinguishes this objective from minimizing

$\min_A MACW(G_A)$00

although every bound on $\min_A MACW(G_A)$01 automatically bounds $\min_A MACW(G_A)$02 because $\min_A MACW(G_A)$03 (Lin et al., 2020).

The proposed approximation algorithm constructs a rooted spanning forest, enumerates $\min_A MACW(G_A)$04 combinations of inter-forest edges when there are $\min_A MACW(G_A)$05 depots, decomposes heavy trees using a lemma of Khani and Salavatipour, and converts each resulting tree into a rooted cycle by doubling edges and shortcutting via the triangle inequality. The final guarantee is a $\min_A MACW(G_A)$06-approximation: $\min_A MACW(G_A)$07 where $\min_A MACW(G_A)$08 is the optimum max-cycle weight. The complexity is

$\min_A MACW(G_A)$09

and for fixed $\min_A MACW(G_A)$10 and $\min_A MACW(G_A)$11 this is $\min_A MACW(G_A)$12 (Lin et al., 2020).

This formulation is sometimes conflated with average-cycle objectives because it controls the worst route among several cycles. The distinction is structural: the paper optimizes maximum total cycle weight, not cycle means, and the presence of depots and cover constraints places it closer to min-max routing than to minimum mean cycle theory.

5. Envy graphs and the named Min Max Average Cycle Weight problem

A paper gives the phrase “The Min Max Average Cycle Weight Problem” a specific meaning in fair allocation. Let $\min_A MACW(G_A)$13 be a set of agents and $\min_A MACW(G_A)$14 a set of objects, with $\min_A MACW(G_A)$15. Each agent $\min_A MACW(G_A)$16 assigns a positive real value $\min_A MACW(G_A)$17 to each object $\min_A MACW(G_A)$18. An allocation $\min_A MACW(G_A)$19 is a perfect matching in the complete bipartite graph $\min_A MACW(G_A)$20, and $\min_A MACW(G_A)$21 denotes the object assigned to agent $\min_A MACW(G_A)$22 (Elmalem et al., 27 Jul 2025).

Given an allocation $\min_A MACW(G_A)$23, the envy of agent $\min_A MACW(G_A)$24 at agent $\min_A MACW(G_A)$25 is

$\min_A MACW(G_A)$26

This defines a complete directed envy graph $\min_A MACW(G_A)$27 on $\min_A MACW(G_A)$28, with arc weights

$\min_A MACW(G_A)$29

For a directed cycle $\min_A MACW(G_A)$30, the average weight is

$\min_A MACW(G_A)$31

and the graph parameter is

$\min_A MACW(G_A)$32

The optimization problem is

$\min_A MACW(G_A)$33

In the clean-slate case, where no preexisting conditions are present, the problem collapses to maximum-value bipartite matching. If

$\min_A MACW(G_A)$34

is maximal over all allocations, then every directed cycle in $\min_A MACW(G_A)$35 has nonpositive total weight, hence $\min_A MACW(G_A)$36. Conversely, if an allocation is not a maximum-value matching, then some directed cycle has positive average weight, so $\min_A MACW(G_A)$37. Therefore an allocation minimizes $\min_A MACW(G_A)$38 if and only if it is a maximum-value matching in $\min_A MACW(G_A)$39, yielding polynomial-time solvability via the assignment problem and, for example, the Hungarian algorithm in $\min_A MACW(G_A)$40 time (Elmalem et al., 27 Jul 2025).

The generalization with preexisting conditions introduces a fixed envy graph $\min_A MACW(G_A)$41 for the original allocation. The new objective is

$\min_A MACW(G_A)$42

where each arc weight in $\min_A MACW(G_A)$43 is

$\min_A MACW(G_A)$44

A concrete $\min_A MACW(G_A)$45 example shows that the allocation minimizing $\min_A MACW(G_A)$46 need not be a maximum-value matching once even a single arc of $\min_A MACW(G_A)$47 has nonzero weight. The corresponding polynomial-time solvability question is left open: $\min_A MACW(G_A)$48 (Elmalem et al., 27 Jul 2025)

This fair-allocation formulation is the most literal instance of a min-max average cycle weight objective in the supplied literature: the inner extremum is a maximum average directed-cycle weight, and the outer extremum is a minimization over allocations.

6. Complexity frontiers and open directions

The fixed-graph problem $\min_A MACW(G_A)$49 is algorithmically well behaved: the fair-allocation paper notes that there exist strongly polynomial-time algorithms for computing $\min_A MACW(G_A)$50 for a given directed graph (Elmalem et al., 27 Jul 2025). The harder part arises when one must optimize over a structured family of graphs, whether over allocations, over cycle covers, or over random graph instances with delicate asymptotics.

For minimum total weight cycle rather than mean cycle, recent distributed work gives a nearly tight approximation tradeoff in the $\min_A MACW(G_A)$51 model. For undirected weighted graphs, there is a randomized $\min_A MACW(G_A)$52-approximation algorithm, for any real $\min_A MACW(G_A)$53, whose round complexity yields a smooth tradeoff between approximation ratio and complexity; when $\min_A MACW(G_A)$54 and $\min_A MACW(G_A)$55, the bound simplifies to

$\min_A MACW(G_A)$56

Assuming the Erdős girth conjecture, any randomized $\min_A MACW(G_A)$57-approximation requires

$\min_A MACW(G_A)$58

rounds even on graphs of diameter $\min_A MACW(G_A)$59 (Chang et al., 26 Mar 2026). These results concern minimum total cycle weight, not average cycle weight, but they delineate the current distributed complexity frontier for cycle-weight approximation.

Two persistent conceptual separations run through the literature. First, total cycle weight, mean cycle weight, and maximum cycle weight within a family are mathematically different objectives, even when they are all described informally as cycle-weight optimization. Second, the most refined asymptotic results are highly model-specific: the threshold $\min_A MACW(G_A)$60 and the $\min_A MACW(G_A)$61 correction belong to the stochastic mean-field complete graph (Ding et al., 2015), whereas the fair-allocation formulation derives its structure from envy graphs and bipartite matchings (Elmalem et al., 27 Jul 2025).

The main unresolved direction in the explicitly named Min Max Average Cycle Weight problem is the preexisting-conditions case $\min_A MACW(G_A)$62, where the clean-slate reduction to maximum-value matching fails (Elmalem et al., 27 Jul 2025). A plausible implication is that future progress will require combining classical cycle-mean computation with optimization over allocations or covers, rather than treating cycle computation and outer optimization as separable tasks.