Evolving combinatorial problem instances that are difficult to solve

被引:35
作者
van Hemert, Jano I. [1 ]
机构
[1] Univ Edinburgh, Natl E Sci Ctr, Edinburgh EH8 9YL, Midlothian, Scotland
关键词
binary constraint satisfaction; travelling salesman; Boolean satisfiability; 3-SAT; difficult combinatorial problems; problem hardness; evolving problems;
D O I
10.1162/evco.2006.14.4.433
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
This paper demonstrates how evolutionary computation can be used to acquire difficult to solve combinatorial problem instances. As a result of this technique, the corresponding algorithms used to solve these instances are stress-tested. The technique is applied in three important domains of combinatorial optimisation, binary constraint satisfaction, Boolean satisfiability, and the travelling salesman problem. The problem instances acquired through this technique are more difficult than the ones found in popular benchmarks. In this paper, these evolved instances are analysed with the aim to explain their difficulty in terms of structural properties, thereby exposing the weaknesses of corresponding algorithms.
引用
收藏
页码:433 / 462
页数:30
相关论文
共 72 条
  • [1] ACHLIOPTAS C, 2000, P 17 NAT C ART INT 1
  • [2] ACHLIOPTAS D, 2004, AAAI, P131
  • [3] ACHLIOPTAS D, 2001, CONSTRAINTS, V4, P329
  • [4] [Anonymous], EFFICIENT CLUSTER CO
  • [5] [Anonymous], THESIS U ALBERTA
  • [6] [Anonymous], 1993, FDN CONSTRAINT SATIS
  • [7] APPLEGATE D, 1999, 99885 RES I DISC MAT
  • [8] APPLEGATE D, 2004, QSOPT LINEAR PROGRAM
  • [9] APPLEGATE D, 2000, CHAINED LIN KERNIGHA
  • [10] Bacchus F, 1998, FIFTEENTH NATIONAL CONFERENCE ON ARTIFICIAL INTELLIGENCE (AAAI-98) AND TENTH CONFERENCE ON INNOVATIVE APPLICATIONS OF ARTIFICAL INTELLIGENCE (IAAI-98) - PROCEEDINGS, P311