Title:A Simple Practical Algorithm for a Computationally Hard Problem

Abstract:Two n-by-n matrices $A$ and $B$ are permutation-similar, if there exists a permutation matrix $P$ such that $A=PBP^t$. Deciding whether two matrices are permutation-similar, is a computationally hard problem. Because, all possible permutations should be checked in general. This decision problem is the core of many main problems, such as graph isomorphism problem. In the graph isomorphism problem, we should decide whether the adjacency matrices of a pair of graphs are permutation-similar. Here, a simple and fast algorithm is proposed to solve this generally hard problem efficiently in practice. The computations show that this proposed simple algorithm which is based on both matrix spectrum and randomness, successfully, solves the graph isomorphism problem for studied instances of hard cases of this problem.