Particle Swarm Optimization (PSO) and two real world applications

Abstract

[en] Particle Swarm Optimization (PSO) belongs to a powerful family of optimization techniques inspired by the collective behavior of social animals. This method has shown promising results in a wide range of applications, especially in computer science. Despite this, a great popularity of such method has not been achieved. Since we believe in the potential of PSO, we propose the following scheme to be able to take advantage of its properties. First, an implementation from scratch in C language of the method has been done, as well as an analysis of its parameters and its performance in function minimization. Then, a second more specific part of this thesis is devoted to the adaptation of the method for solving two real-world applications. The first one, in the field of signal analysis, consists of an optimization method for the numerical analysis of Fourier functions, whereas the second, in the field of computer science, comprises the optimization of neural networks weights’ for some small architectures.