Multi-Task Code Reuse in Genetic Programming | dr hab. Wojciech Jaśkowski

by Wojciech Jaśkowski, Krzysztof Krawiec, Bartosz Wieloch

Abstract:

We propose a method of knowledge reuse between evolutionary processes that solve different optimization tasks. We define the method in the framework of tree-based genetic programming (GP) and implement it as code reuse between GP trees that evolve in parallel in separate populations delegated to particular tasks. The technical means of code reuse is a crossbreeding operator which works very similar to standard tree-swapping crossover. We consider two variants of this operator, which differ in the way they handle the incompatibility of terminals between the considered problems. In the experimental part we demonstrate that such code reuse is usually benefficial and leads to success rate improvements when solving the common boolean benchmarks.

View PDF

Reference:

Multi-Task Code Reuse in Genetic Programming (Wojciech Jaśkowski, Krzysztof Krawiec, Bartosz Wieloch), In GECCO-2008 Late-Breaking Papers (Marc Ebner, Mike Cattolico, Jano van Hemert, Steven Gustafson, Laurence D. Merkle, Frank W. Moore, Clare Bates Congdon, Christopher D. Clack, Frank W. Moore, William Rand, Sevan G. Ficici, Rick Riolo, Jaume Bacardit, Ester Bernado-Mansilla, Martin V. Butz, Stephen L. Smith, Stefano Cagnoni, Mark Hauschild, Martin Pelikan, Kumara Sastry, eds.), Association for Computing Machinery, 2008.

Bibtex Entry:

@InProceedings{Jaskowski2008multi-task,
  Title                    = {Multi-Task Code Reuse in Genetic Programming},
  Author                   = {Wojciech Jaśkowski and Krzysztof Krawiec and Bartosz Wieloch},
  Booktitle                = {GECCO-2008 Late-Breaking Papers},
  Year                     = {2008},

  Address                  = {Atlanta, GA, USA},
  Editor                   = {Marc Ebner and Mike Cattolico and Jano {van Hemert} and Steven Gustafson and Laurence D. Merkle and Frank W. Moore and Clare Bates Congdon and Christopher D. Clack and Frank W. Moore and William Rand and Sevan G. Ficici and Rick Riolo and Jaume Bacardit and Ester Bernado-Mansilla and Martin V. Butz and Stephen L. Smith and Stefano Cagnoni and Mark Hauschild and Martin Pelikan and Kumara Sastry},
  Month                    = {12-16 } # jul,
  Organization             = {Association for Computing Machinery},
  Pages                    = {2159--2164},
  Publisher                = {Association for Computing Machinery},

  Abstract                 = {We propose a method of knowledge reuse between evolutionary processes that solve different optimization tasks. We define the method in the framework of tree-based genetic programming (GP) and implement it as code reuse between GP trees that evolve in parallel in separate populations delegated to particular tasks. The technical means of code reuse is a crossbreeding operator which works very similar to standard tree-swapping crossover. We consider two variants of this operator, which differ in the way they handle the incompatibility of terminals between the considered problems. In the experimental part we demonstrate that such code reuse is usually benefficial and leads to success rate improvements when solving the common boolean benchmarks.},
  ISBN                     = {978-1-60558-131-6},
  Keywords                 = {genetic algorithms, genetic programming, code Reuse, multi-task learning},
  Publisher_address        = {New York, NY, USA},
  Url                      = {http://www.cs.bham.ac.uk/~wbl/biblio/gecco2008/docs/p2159.pdf}
}