Scaling algorithms for approximate and exact maximum weight matching

The {\em maximum cardinality} and {\em maximum weight matching} problems can be solved in time $\tilde{O}(m\sqrt{n})$, a bound that has resisted improvement despite decades of research. (Here $m$ and $n$ are the number of edges and vertices.) In this article we demonstrate that this "$m\sqrt{n}$ barrier" is extremely fragile, in the following sense. For any $\epsilon>0$, we give an algorithm that computes a $(1-\epsilon)$-approximate maximum weight matching in $O(m\epsilon^{-1}\log\epsilon^{-1})$ time, that is, optimal {\em linear time} for any fixed $\epsilon$. Our algorithm is dramatically simpler than the best exact maximum weight matching algorithms on general graphs and should be appealing in all applications that can tolerate a negligible relative error. Our second contribution is a new {\em exact} maximum weight matching algorithm for integer-weighted bipartite graphs that runs in time $O(m\sqrt{n}\log N)$. This improves on the $O(Nm\sqrt{n})$-time and $O(m\sqrt{n}\log(nN))$-time algorithms known since the mid 1980s, for $1\ll \log N \ll \log n$. Here $N$ is the maximum integer edge weight.