Current Proceedings on Technology
Yazarlar: Carlos Cernuda, Edwin Lughofer, Wolfgang Märzinger, Wolfram Summerer
Konular:-
Anahtar Kelimeler:Differential evolution,Particle swarm optimization,Ant colony optimization,Genetic algorithm
Özet: Nowadays, the techniques employed in data acquisition in Chemometrics (e.g. NIR or MIR) can provide huge amounts of data in a cheap way. Thus, a tsunami of data, where the number of variables explodes, must be employed, being necessary a variable selection approach as a previous step in any classification or regression problem. Particle Swarm Optimization (PSO) and Ant Colony Optimization (ACO) are searching algorithms that have been recently used for that purpose. Due to the nature of the search procedures, both suffer from the problem of being trapped in local optima that could differ much from the global ones, which are unknown. In the line of Differential Evolution, the authors propose a hybridization of both algorithms by means of a Genetic Algorithm (GA) approach, which combines the advantages of both searching algorithms and promotes escaping from local optima.