ANALYSIS OF THE HELD-KARP HEURISTIC FOR THE TRAVELING SALESMAN PROBLEM

  • Authors:
  • D. P. Williamson

  • Affiliations:
  • -

  • Venue:
  • ANALYSIS OF THE HELD-KARP HEURISTIC FOR THE TRAVELING SALESMAN PROBLEM
  • Year:
  • 1990

Quantified Score

Hi-index 0.00

Visualization

Abstract

The Held-Karp heuristic for the Traveling Salesman Problem (TSP) has in practice provided near-optimal lower bounds on the cost of solutions to the TSP. We analyze the structure of Held-Karp solutions in order to shed light on their quality. In the symmetric case with triangle inequality, we show that a class of instances has planar solutions. We also show that Held-Karp solutions have a certain monotonicity property. This leads to an alternate proof of a result of Wolsey, which shows that the value of Held-Karp heuristic is always at least 2/3 OPT, where OPT is the cost of the optimum TSP tour. Additionally, we show that the value of the Held-Karp heuristic is equal to that of the linear relaxation of the biconnected-graph problem when edge costs are non-negative. In the asymmetric case with triangle inequality, we show that there are many equivalent definitions of the Held-Karp heuristic, which include finding optimally weighted 1-arborescences, 1-antiarborescences, asymmetric 1-trees, and assignment problems. We prove that monotonicity holds in the asymmetric case as well. These theorems imply that the value of the Held-Karp heuristic is no less than OPT and no less than the value of the Balas-Christofides heuristic for the asymmetric TSP. For the 1,2-TSP, we show that the Held-Karp heuristic cannot do any better than 9/10 OPT, even as the number of nodes tends to infinity. Portions of this thesis are joint work with David Shmoys.