This article was accepted into the corpus but its outbound wikilinks were never NER-processed — typical at the deepest BFS hop or when the run's entity cap was reached. No expansion funnel to show.
| Stoer–Wagner algorithm | |
|---|---|
| Name | Stoer–Wagner algorithm |
| Author | Mechthild Stoer, Frank Wagner |
| Introduced | 1990 |
| Genre | Graph algorithm |
| Problem | Global minimum cut |
| Time complexity | O(n m + n^2 log n) (original); improvements possible |
Stoer–Wagner algorithm
The Stoer–Wagner algorithm is a deterministic algorithm for computing a global minimum cut in an undirected weighted graph. It was introduced by Mechthild Stoer and Frank Wagner and has been influential in combinatorial optimization and network analysis. The algorithm repeatedly identifies minimum s–t cuts via maximum-adjacent-vertex selection and contracts vertices to produce a global minimum cut.
The algorithm was published by Mechthild Stoer and Frank Wagner in 1990 as part of the literature on graph algorithms and combinatorial optimization. Its development followed foundational work on network flows by L. R. Ford Jr. and D. R. Fulkerson and is contemporaneous with advances by Jack Edmonds and Richard Karp in polynomial-time combinatorial methods. Subsequent algorithmic research by researchers such as Robert Tarjan, Michael Stoer (note: distinct contributions), and Karger, David led to randomized alternatives and refinements that connected to the work of Karger, Clifford on contraction algorithms. The Stoer–Wagner method influenced practical developments in software libraries maintained by institutions such as AT&T. Its theoretical context relates to results by Nagamochi, Hideo and Iwata, Satoru on connectivity and cut packing.
Given an undirected weighted graph G = (V, E, w) the goal is to find a nonempty proper subset S ⊂ V that minimizes the total weight of edges crossing between S and V \ S. This global minimum cut problem generalizes unweighted min-cut problems studied in literature including works on the Max-Flow Min-Cut Theorem by L. R. Ford Jr. and D. R. Fulkerson and complements s–t min-cut formulations used in studies by John Hopcroft and Robert Tarjan. The result is a partition of vertices whose cut weight is minimal among all possible partitions, which has implications in contexts addressed by researchers at institutions like Bell Labs, MIT, and IBM Research.
Stoer–Wagner repeatedly performs n−1 phases on graph G. Each phase grows a set A by repeatedly adding the most tightly connected vertex v ∈ V\A, measured by summed incident edge weights to A, reminiscent of greedy selections in methods by Edsger Dijkstra (for shortest paths) and greedy heuristics used in work by Jack Edmonds. After A grows to include all vertices, the last added vertex t and the previously added vertex s define an s–t cut; the algorithm records the cut weight and then contracts s and t into a single vertex, updating incident weights, similar in spirit to contraction steps studied by David Karger. The minimum recorded cut across all phases is returned. Implementation often uses priority data structures such as binary or Fibonacci heaps linked to innovations by Michael Fredman and Robert Tarjan to manage selection of the maximum-weight vertex efficiently.
The original Stoer–Wagner implementation runs in O(n m + n^2 log n) time using priority queues, where n = |V| and m = |E|. With adjacency structures and optimized data handling, performance matches or improves over alternative deterministic algorithms by researchers like Nagamochi, Hideo and Takahashi, Hisao. Randomized algorithms by David Karger achieve expected near-linear performance in certain regimes, which contrasts with Stoer–Wagner’s deterministic guarantees favored in industry settings at Google and Microsoft Research for predictable runtime. Practical implementations in graph libraries such as those from Boost (C++ Libraries) and scientific computing groups at Stanford University and University of California, Berkeley exploit sparsity and engineering optimizations to reduce constant factors.
Correctness rests on the property that each phase yields a minimum s–t cut over the particular choice of s and t produced by the greedy vertex-addition sequence; contracting s and t preserves minimum global cut weight. Proofs employ induction over phases and cut-splitting arguments that relate to submodularity properties studied by László Lovász and duality principles tracing back to the Max-Flow Min-Cut Theorem by L. R. Ford Jr. and D. R. Fulkerson. Formal analyses reference matroid and cut-set concepts appearing in works by William Tutte and combinatorial frameworks developed by Jack Edmonds. Alternative correctness treatments reduce to cut and contraction invariants used in randomized contraction proofs by David Karger.
Several deterministic and randomized variants have been proposed. Nagamochi–Ibaraki algorithms by Nagamochi, Hideo and T. Ibaraki provide alternative deterministic sparse-certificate approaches. Karger’s randomized contraction algorithm by David Karger and later improvements by David Karger and collaborators yield Monte Carlo variants with high-probability guarantees. Extensions integrate stoer–Wagner-like phases into global optimization frameworks studied at Bell Labs and in clustering pipelines at University of Oxford. Parallel and distributed adaptations have been explored in contexts such as high-performance computing centers at Argonne National Laboratory and cloud infrastructures by Amazon Web Services and Google Cloud Platform.
The Stoer–Wagner algorithm is applied in network reliability assessments studied by AT&T Bell Laboratories, image segmentation research influenced by groups at ETH Zurich and University of Toronto, and community-detection tasks pursued by teams at Facebook and LinkedIn. It supports preprocessing in flow-based methods used in computational biology projects at Broad Institute and in circuit partitioning work from Intel and Cadence Design Systems. The algorithm’s deterministic nature makes it suitable for regulatory and infrastructure planning problems addressed by organizations like National Science Foundation-funded research groups and industry labs at IBM Research.