Abstract
Counting perfect matchings has played a central role in the theory of
counting problems. The permanent, corresponding to bipartite graphs, was shown
to be #P-complete to compute exactly by Valiant (1979), and a fully polynomial
randomized approximation scheme (FPRAS) was presented by Jerrum, Sinclair, and
Vigoda (2004) using a Markov chain Monte Carlo (MCMC) approach. However, it has
remained an open question whether there exists an FPRAS for counting perfect
matchings in general graphs. In fact, it was unresolved whether the same Markov
chain defined by JSV is rapidly mixing in general. In this paper, we show that
it is not. We prove torpid mixing for any weighting scheme on hole patterns in
the JSV chain. As a first step toward overcoming this obstacle, we introduce a
new algorithm for counting matchings based on the Gallai-Edmonds decomposition
of a graph, and give an FPRAS for counting matchings in graphs that are
sufficiently close to bipartite. In particular, we obtain a fixed-parameter
tractable algorithm for counting matchings in general graphs, parameterized by
the greatest "order" of a factor-critical subgraph.