Comparison of Metaheuristics

We propose a simple experiment to test this hypothesis

John Silberholz


Scholarcy highlights

  • Metaheuristics are truly diverse in nature — under the overarching theme of performing operations to escape local optima, algorithms as different as ant colony optimization, and genetic algorithms () have emerged
  • We believe following the procedures described in this paper will increase the quality of metaheuristic comparisons
  • Choosing an appropriate testbed and distributing it so other researchers can access it will result in more high-quality comparisons of metaheuristics, as researchers will test on the same problem instances
  • Expanding the practice of creating geometric problem instances with easy-to-visualize optimal or near-optimal solutions will increase understanding of how metaheuristics perform in a global optimization sense
  • It is important to recognize that the number of algorithm parameters has a direct effect on the complexity of the algorithm and on the number of parameter interactions, which complicates analysis
  • If the number of parameters is considered in the analysis of metaheuristics, this will encourage simpler, easier-toanalyze procedures

Need more features? Save interactive summary cards to your Scholarcy Library.