Sparse Matrices Reordering using Evolutionary Algorithms: A Seeded Approach

Abstract

In this work, it is introduced a methodology for solving the problem of sparse matrices reordering using evolutionary algorithms, which can be handled as a combinatorial NP-class problem. Evolutionary algorithms are more flexible techniques that allow this reordering considering location and also values of the non zero entries of the matrix. Different fitness functions are proposed and studied comparatively. Moreover, the obtained results are compared with a classical procedure, the inverse Cuthill-McKee ordering. Finally, a seeded approach that combines both strategies, whose results outperform the previous ones, is introduced