PARTICLE SWARM OPTIMIZATION

Discrete Particle Swarm Optimization for Ontology Alignment



Discrete Particle Swarm Optimization for Ontology Alignment

Introduction

PSO (Particle Swarm Optimization) is an optimization meta-heuristic which is biologically-inspired which gain gained momentum continuously in recent years. This technique can be applied generally to problems in multi-dimensional search space where objective function requires its global optimum. Although particle swarm optimization was originally developed to solve continuous optimization problems, number of modifications proposed by experts in the field can potentially make it its application possible in discrete optimization problems. Run-time behavior and convergence of particle swarm optimization cannot be determined trivially, however, numerous applications and case studies have increasingly showed that this technique is more reliable than the techniques of non-heuristic optimizers and performs better usually. Development of intelligent system holds the need for knowledge and information to be presented in a form which machines can read. In computer science context, ontologies are described formally by Gruber (1993). He defined it formally as a specification of conceptualization which is explicit. This definition has refined frequently which causes its reinterpretation in numerous contexts.

Particle swarm algorithm is relatively recent, but various researchers have suggested a number of modifications and new works on this subject do not cease to be published. There are several ways to improve the classical algorithm, implemented in most of them. This compound is an algorithm with other algorithms, optimization, reducing the likelihood of premature convergence by changing the characteristics of particle motion and the dynamic change of the parameters of the algorithm during the optimization. 

Research Problem

Particle Swarm Optimization (PSO) is a biologically-inspired, population-based optimization technique that has been successfully applied to various problems in science and engineering. This paper will study a new angle of the discrete particle swarm optimization for ontology alignment.

Literature Review

Alignment can be defined in terms of set of correspondence among/between the ontological entities (properties, classes) and different individuals from each ontologies. It is understandable that figuring out the best unique alignment for two ontologies is quite difficult task, which without a doubt is an inherently hard or even next to impossible to achieve. Its application is not only limited to ontology automatic alignment system as it can be applied to humans as well. Hence, optimal alignment is not known quiet often as there is an absence of any 'gold standard' which can be used as a reference (Ehrig & Euzenat, 2005, pp. 25). Another major problem in this regard is related to the size of ontologies which is increasing.

For example, in the domain of medical, in cases related to library use, or in other thesauri which is large scale in which ontologies with hundreds of thousands of concepts are common. In this paper, ontology alignment problem are regarded as optimization problem and adapted discrete Particle Swarm Optimization (DPSO) algorithm is applied to figure out the optimal alignment. Application of Particle Swarm Optimization to ontology alignment problem offers several potential benefits. Firstly, huge amount of inputs can be processed easily as iterative traversal related to large ontologies can be ...