Parallel Multi-Swarm PSO Based on K-Medoids and Uniform Design

Abstract

PAM (Partitioning around Medoid) is introduced to divide the swarm into several different subpopulations. PAM is one of k-medoids clustering algorithms based on partitioning methods. It attempts to divide n objects into k partitions. This algorithm overcomes the drawbacks of being sensitive to the initial partitions in kmeans algorithm. In the parallel PSO algorithms, the swarm needs to be divided into several different smaller swarms. This study can be excellently completed by PAM. The aim of clustering is that particles within the same sub-population are relative concentrative, so that they can be relatively easy to learn. The purposes of this strategy are that the limited time will be spent on the most effective search; therefore, the search efficiency can also be significantly improved. In order to explore the whole solution space evenly, uniform design is introduced to generate an initial population, in which the population members are scattered uniformly over the feasible solution space. In evolution, uniform design is also introduced to replace some worse individuals. Based on abovementioned these technologies, a novel algorithm, parallel multi-swarm PSO based on k-medoids and uniform design, is proposed. A difference between the proposed algorithm and the others is that PAM and uniform design are both firstly introduced to parallel PSO algorithms.

A Particle Swarm Optimization Algorithm Based on Uniform Design

“…The core idea of uniform design is to scatter design points uniformly on an experimental region [2,8,9], it was proposed by Professor Kaitai Fang and Yuan Wang [5][6][7]. UD has been successfully used in many fields, such as computer sciences, system engineering, quality engineering, pharmaceutics, survey design, and natural sciences, etc.…”

“…Particle Swarm Optimization Algorithm (PSO) is one of swarm intelligence optimization algorithms [2], it is belongs to random searching algorithms [1]. In 1995 Eberhart and Kennedy proposed PSO algorithm.…”

“…The path of particle is updated through individual best position and the path of swarm is updated via global best location, which is found by the entire population. This makes particles move to the optimal solution [2]. The initial population and the initial value of algorithm parameters in PSO method, have direct influence on the parameter results identification.…”

“…This algorithm thinks about the uniformity of the particle swarm distribution and the memory characteristic of particles. In the reference [2], is proposed a novel algorithm, parallel multi-swarm PSO based on k-medoids and uniform design. This paper proposed a new Particle Swarm Optimization Algorithm based on Uniform Design.…”

Attribute Index and Uniform Design Based Multiobjective Association Rule Mining with Evolutionary Algorithm

“…The main objective of uniform design is to sample a small set of points from a given set of points, such that the sampled points are uniformly scattered [23, 33–35]. …”