Skip to content

Benchmarking MWM in Complete Bigraphs

Jared Beck edited this page Jan 18, 2015 · 3 revisions

Uses the Augmenting Path algorithm from Maximum Cardinality Matching, with the "scaling" approach described by Gabow (1983) and Galil (1986).

MWM in Complete Bigraph is O(n ** (3/4) m log N)

The bigraphs used in this benchmark are:

  1. Complete - Each vertex in U has an edge to each vertex in V
  2. Weighted - Each edge has a different, consecutive integer weight
  3. "Balanced"
  • In an even bigraph, U and V have the same number of vertexes.
  • In an odd bigraph, U has one more than V.
  • 0 <= abs(U - V) <= 1

These attributes produce the "worst case" bigraph, i.e. the most expensive over which to produce a complete matching.

Plot generated using gnuplot. Benchmark scripts and data may be found in /benchmark.