AutoML.org

Freiburg-Hannover-Tübingen

ParamILS

ParamILS is a well-performing method for parameter tuning and algorithm configuration using iterated local search.

It has been used in dozens of academic applications to improve state-of-the-art solvers for more then ten hard computational problems (see its 250+ Google scholar citations). ParamILS has also had an impact in several industrial applications. In all these applications, ParamILS has yielded substantial speedups of state-of-the-art solvers for hard combinatorial problems, such as propositional satisfiability (SAT), mixed integer programming (MIP), AI planning, answer set programing (ASP), and timetabling. By doing so, ParamILS has helped several systems win solver competitions.

 Website (external)

References

  • ParamILS: An Automatic Algorithm Configuration Framework
    Frank Hutter, Holger H. Hoos, Kevin Leyton-Brown, and Thomas Stützle. Journal of Artificial Research 2009, volume 36, pages 267-306.
  • Automatic Algorithm Configuration based on Local Search
    Frank Hutter, Holger Hoos, and Thomas Stützle – Proc. of AAAI 2007.
  • Parallel algorithm configuration
    Frank Hutter, Holger H. Hoos, and Kevin Leyton-Brown. Proc. of LION 2012.