Parallel genetic algorithm based on Thread-Level Speculation

Abstract

An improved method of classic genetic algorithms is proposed which uses Thread-Level Speculation (TLS) technology for the shortcoming that classic genetic algorithm's search speed is slow. Unlike the classic genetic algorithm, Firstly we replace an individual with a single thread, use TLS technology to process each thread parallel, in order to eliminate the correlation among threads, increase individual diversity in the population, thus accelerate the convergence speed of genetic algorithms. Experimental results demonstrate that the method can increase the individual diversity in the population and accelerate the convergence speed of genetic algorithms, compared with the classic genetic algorithms.