The matching problem for bipartite graphs with polynomially bounded permanents is in NC

Abstract

It is shown that the problem of deciding and constructing a perfect matching in bipartite graphs G with the polynomial permanents of their n × n adjacency matrices A (perm(A) = nO(1)) are in the deterministic classes NC2 and NC3, respectively. We further design an NC3 algorithm for the problem of constructing all perfect matchings (enumeration problem) in a graph G with a permanent bounded by O(nk). The basic step was the development of a new symmetric functions method for the decision algorithm and the new parallel technique for the matching enumerator problem. The enumerator algorithm works in O(log3 n) parallel time and O(n3k+5.5 ¿ log n) processors. In the case of arbitrary bipartite graphs it yields an 'optimal' (up to the log n- factor) parallel time algorithm for enumerating all the perfect matchings in a graph. It entails also among other things an efficient NC3-algorithm for computing small (polynomially bounded) arithmetic permanents, and a sublinear parallel time algorithm for enumerating all the perfect matchings in graphs with permanents up to 2nε.