Improved Smoothed Analysis of 2-Opt for the Euclidean TSP

arXiv2211.16908MaRDI QIDQ6419040FDOQ6419040

Authors: Bodo Manthey, Jesse van Rhijn 

Publication date: 30 November 2022

Abstract: The 2-opt heuristic is a simple local search heuristic for the Travelling Salesperson Problem (TSP). Although it usually performs well in practice, its worst-case running time is poor. Attempts to reconcile this difference have used smoothed analysis, in which adversarial instances are perturbed probabilistically. We are interested in the classical model of smoothed analysis for the Euclidean TSP, in which the perturbations are Gaussian. This model was previously used by Manthey & Veenstra, who obtained smoothed complexity bounds polynomial in $n$, the dimension $d$, and the perturbation strength $sigm{a}^{-1}$. However, their analysis only works for $dgeq4$. The only previous analysis for $dleq3$ was performed by Englert, R"oglin & V"ocking, who used a different perturbation model which can be translated to Gaussian perturbations. Their model yields bounds polynomial in $n$ and $sigm{a}^{-d}$, and super-exponential in $d$. As no direct analysis existed for Gaussian perturbations that yields polynomial bounds for all $d$, we perform this missing analysis. Along the way, we improve all existing smoothed complexity bounds for Euclidean 2-opt.

This page was built for publication: Improved Smoothed Analysis of 2-Opt for the Euclidean TSP

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6419040)