2026-10-29T21:49:37Zhttps://munzur-rdlopen.rawdatalibrary.net/oai/request

oai:munzur-rdlopen.rawdatalibrary.net:munzur/43392026-10-28T15:31:29Zopenairedrivercom_munzur_1col_munzur_2com_munzur_8col_munzur_157

A Consensus-Edge Initializer for Multi-start 2-opt on the Symmetric Euclidean Travelling Salesman ProblemSaygili, Ahmet CaglarKazan, HalimTravelling Salesman Problemlocal search-optconsensus edgesheuristicsWe study a one-shot consensus-based initializer for multi-start 2-opt on symmetric Euclidean Travelling Salesman Problem (TSP) instances. Given the M local minima from best-of-M 2-opt, we count undirected edge frequencies, greedily assemble a tour from the most frequent edges with a length-biased tie-break, and run a single additional 2-opt descent from this seed. The local search operator is unchanged; the consensus stage reuses existingtours and adds only a small overhead beyond the M descents, and we characterise this overhead analytically while focusing evaluation on tour length improvements. On synthetic Euclidean instances from three generators, we evaluate performance over a grid of problem sizes n and restart budgets M. The consensus + 2-opt pipeline improves best-of-M 2-opt by roughly 3-6% in median tour length with high win rates for n >= 500 and moderate M, while for small n and large M the baseline saturates and gains shrink to zero or slightly negative values. Thus, even a single aggregation step over 2-opt local minima yields a simple, reproducible seed that is beneficial in the practically relevant regime of moderate budgets and larger instances and approximately neutral elsewhere.ISTANBUL UNIV2026Articlehttps://rdlopen.rawdatalibrary.net/r/munzur/4339https://doi.org/10.26650/ekoist.2026.xx.18248642651-396XenEKOIST-JOURNAL OF ECONOMETRICS AND STATISTICSinfo:eu-repo/semantics/openAccessinfo:eu-repo/semantics/articlehttps://munzur-rdlopen.rawdatalibrary.net/tr/items/munzur/4339