Evolutionary Algorithm Based Function Optimization with Solution Structure Analysis

Problem optimizacije sveprisutan je u području računarske znanosti kao i u mnogim drugim područjima. Optimizacijski problem može biti definiran nad diskretnom domenom, kontinuiranom domenom ili mješavinom obaju domena. Ovisno o domeni problema, razvijeno je mnoštvo različitih pristupa za njihovo rješavanje. Dimenzionalnost domene pri tome predstavlja ograničavajući faktor, pri čemu probleme s visoko-dimenzijskim domenama često nije moguće rješavati učinkovito. Jedan od pristupa koji bi kod određenih vrsta problema mogao rezultirati povećanjem učinkovitosti pronalaska kvalitetnih rješenja jest dinamičko smanjenje dimenzionalnosti problema na način da se pojedine komponente domene grupiraju. Primjer takvog pristupa opisan je u radu "Transforming Evolutionary Search into Higher-Level Evolutionary Search by Capturing Problem Structure", IEEE Trans. on Evolutionary Computation, October 2014, Vol 18, No 5.The optimization problem is ubiquitous in the field of computer science as well as in many other areas. The optimization problem can be defined over a discrete domain, a continuous domain, or a mixture of both domain. Depending on the domain of the problem, a multitude of different approaches have been developed for solving them. Domain dimensionality represents a limiting factor, where problems with high-dimensional domains can often not be effectively addressed. One approach that could result in certain types of problems increasing the efficiency of finding quality solutions is dynamic reducing the dimensionality of the problem in a way that individual components domains are grouped. An example of such approach is described in "Transforming evolutionary Search in Higher-Level Evolutionary Search by Capturing problem Structure ", IEEE Trans. On Evolutionary Computation, October 2014, Vol 18, No 5