摘要

For significant performance enhancement of GA, we introduce a new and simple operation with the name Trans-addition involving binary addition of selected chromosome to Pseudo-blank chromosomes. A malePseudo-blank has all but one of its bits set to zero. Its 2*s complement is the female. Trans-addition forces smallest possible change to each decision variable. The probability of such a change in canonical GA is quite low. Only the fittest chromosome performs Trans-addition with all Pseudo-blanks. It satisfies a natural behavior of higher opportunities of reproduction for the fittest. Trans-addition does not interfere with regular GA operations but only supplements. The implementation, named ※Trans-GA§, is based on the representation of real-valued decision variables in terms of variable integer increments and user-specified accuracy. Our results clearly prove that Trans-GA outperforms the canonical GA in reliability and convergence rate by a wide margin. The presented GA implementation suits Engineering Design Optimization problems involving integer, real and mixed decision variables.

  • 出版日期2010

全文