Embracing Plagiarism: Theoretical, Biological and Empirical Justification for Copy Operators in Genetic Optimisation

摘要

A novel genetic operator, the plagiarism operator, is introduced for evolutionary design and optimisation. This operator is analogous in some respects to crossover and to biological transposition. Plagiarism is shown to be theoretically superior to uniform mutation for generalised counting-ones problems, and also to outperform uniform mutation on certain classes of random fitness landscapes. Experimental results are presented showing that plagiarism speeds up the artificial evolution of certain digital logic circuits. The performance of this operator is interpreted in terms of the non-uniform distribution of genetic primitives in good solutions for certain problems.