Particle swarm optimization algorithm with variable random function

The random function of standard particle swarm optimization algorithm is distributed uniformly on the range of [0, 1], which makes the algorithm become a stochastic search method. To avoid the trap into blindness of stochastic search, research on the distribution of random function is discussed. Unlike the traditional direction of search based on gradient, making the best use of thought of statistics, a new direction is proposed, based on the concept of skewness. Selecting an appropriate random function to adapt to skewness, particle swarm optimization algorithm with variable random function (VRFPSO) is proposed. Using the 4 common Benchmark functions, the experiment results show that it has a better performance in global search capacity and convergence rate. In retrospect, the results approve the significant effect of distribution of random function on particle swarm optimization and research on new direction of search based on statistical information is of great importance.