Leveraging TSP Solver Complementarity through Machine Learning

Pascal Kerschke, Lars Kotthoff, Jakob Bossek, Holger H. Hoos, Heike Trautmann
2017 Evolutionary Computation  
The Travelling Salesperson Problem (TSP) is one of the best-studied NP-hard problems. Over the years, many different solution approaches and solvers have been developed. For the first time, we directly compare five state-of-the-art inexact solversnamely, LKH, EAX, restart variants of those, and MAOS-on a large set of well-known benchmark instances and demonstrate complementary performance, in that different instances may be solved most effectively by different algorithms. We leverage this
more » ... mentarity to build an algorithm selector, which selects the best TSP solver on a perinstance basis and thus achieves significantly improved performance compared to the single best solver, representing an advance in the state of the art in solving the Euclidean TSP. Our in-depth analysis of the selectors provides insight into what drives this performance improvement. Evolutionary Computation 26 (4) : 597-620 UBC (cheap). A subset of 13 computationally cheap features from UBC feature set by Hutter et al. (2014), excluding local search, branch and cut, and clustering distance features. 5 https://CRAN.R-project.org/package=tspmeta 6
doi:10.1162/evco_a_00215 pmid:28836836 fatcat:5awgbc3sfngohgr6gq243v3vlm