This paper proposes a new type of mutation operator, FEDS (Fitness, Elitism, Depth, and Size) mutation in genetic programming. The concept behind the new mutation operator is inspired from already introduced FEDS crossover operator to handle the problem of code bloating. FEDS mutation operates by using local elitism replacement in combination with depth limit and size of the trees to reduce bloat with a subsequent improvement in the performance of trees (program structures). We have designed a multiclass classifier for some benchmark datasets to test the performance of proposed mutation. The results show that when the initial run uses FEDS crossover and the concluding run uses FEDS mutation, then not only is the final result significantly improved but there is reduction in bloat also.
Anuradha Purohit1, Narendra S. Choudhari2 and Aruna Tiwari3
1Shri Govindram Seksaria Institute of Technology and Science, India,2,3Indian Institute of Technology Indore, India
Bloat, Crossover, Elitism, Fitness, Mutation, Reproduction | Published By :
ICTACT
Published In :
ICTACT Journal on Soft Computing
( Volume: 3 , Issue: 2 )
Date of Publication :
January 2013
|