Low-cost heuristics for matrix bandwidth reduction combined with a Hill-Climbing strategy

This paper studies heuristics for the bandwidth reduction of large-scale matrices in serial computations. Bandwidth optimization is a demanding subject large number scientific and engineering applications. A heuristic labels rows columns given sparse matrix. The algorithm arranges entries with nonzero coefficient as close to main diagonal possible. modifies an ant colony hyper-heuristic approach generate expert-level combined Hill-Climbing strategy when applied arising from specific application areas. Specifically, this uses low-cost state-of-the-art tandem procedure. results yielded on wide-ranging set standard benchmark showed that proposed outperformed symmetric sparsity patterns.