Parallelization of Elimination Tournament Selection without Synchronization

 A global parallel genetic algorithm (GPGA) with elimination tournament selection without duplicates is described in this paper. The parallel implementation of genetic algorithm with multiple threads without synchronization is thoroughly investigated. The problem which occurs when several threads select the same individual for elimination is recognized in such implementation. The probability of that event is analytically formulated. Considering that probability, we determine the total number of iterations the asynchronous GPGA should perform in order to achieve the same optimization effect as a serial GA or a synchronous GPGA.