AMCSE 2017

The 2017 International Conference Applied Mathematics, Computational Science and Systems Engineering
Athens, Greece, October 6-8, 2017


The conference will be held in a Sea Resort Hotel 40 Kms outside Athens and 30 Kms away from the Athens International Airport.



PLENARY SPEAKERS of 2017

 


Prof. Ivan Zelinka, Faculty of Electrical Engineering and Computer Science, VŠB-TUO Ostrava, Czech Republic, ivan.zelinka@vsb.cz

Title: "Novel Approach in Evolutionary Algorithms Dynamics Analysis and Control Recent Advances and Progress"

Abstract:

This tutorial is focused on recent progress in novel approach on evolutionary algorithm dynamics analysis and control. Based on our research and research of various scientists worldwide it has been shown and experimentally demonstrated on many evolutionary algorithms, that its dynamics is equivalent or similar to dynamics of complex/social networks, see Fig. 1. Evolutionary algorithms can be then visualized, analyzed and controlled in this way, i.e. by means of already existing methods for complex networks. As already reported in many research papers and books, e.g. [1]-[11] the dynamics can generate different kinds of behavior including chaotic one and can be visualized as a complex geometrical structure. In the tutorial, there will be explained relations between evolutionary dynamics, its visualization as a complex network and actual state of our novel methods of its analysis and control. Proposed approach will be demonstrated on already published results (see for example [1]-[11]) with selected evolutionary algorithms as for example differential evolution, genetic algorithm, particle swarm, artificial bee algorithm, and others as ACO is. Relation between complex networks attributes, including their time dependence, and evolutionary dynamics will be explained. Methodology converting evolutionary algorithms to the complex network will be introduced including demonstrations in Mathematica software (free to download and use). Tutorial will then continue by explanation how evolutionary dynamics can be successfully controlled in order to improve performance of given algorithm, as already reported for example in [1]-[11]. At the end, based on this idea of control, we show how EA can be understood as feedback loop control system.


Fig. 1 Individual interactions in evolutionary algorithm (left), on arbitrary problem, can be captured like social interaction and thus social-like network is created (right). Edges then exhibit an importance of individual in the population in the time.




 



Contact us

amcse.conference@gmail.com
bardis@ieee.org
bardis@ilabsse.gr

Categories

Academic Sponsors