Samoprilagajanje krmilnih parametrov pri algoritmu diferencialne evolucije za večkriterijsko optimizacijo

magistrsko delo

Aleš Zamuda (Author), Janez Brest (Mentor)

Abstract

Predstavljamo nov algoritem diferencialne evolucije za večkriterijsko optimizacijo, ki ga krmili samoprilagoditveni mehanizem, predstavljen v evolucijskih strategijah. Posebej se posvetimo načrtu algoritma, tj. matematično formalnemu zapisu glavnih sestavnih delov algoritma in medsebojni povezavi teh delov. Potek algoritma opišemo s psevdokodom. Sklepamo na časovno zahtevnost algoritma in podamo nekaj empiričnih meritev, da bi jo potrdili. Posebej preučimo tudi dinamiko samo-prilagajanja krmilnih parametrov. Navedemo sodobne testne probleme in indikatorje kakovosti iz literature za oceno zmogljivosti večkriterijskih optimizacijskih algoritmov. S pomočjo teh dobimo ocene zmogljivosti algoritma, ki pokažejo številna statistično signifikantna izboljšanja. Dobljene rezultate našega algoritma primerjamo tudi s sorodnimi algoritmi in na empiričnih rezultatih pokažemo, kje je algoritem statistično signifikantno boljši.

Keywords

evolucijski algoritmi;diferencialna evolucija;samoprilagajanje krmilnih parametrov;večkriterijska optimizacija;

Data

Language:	Slovenian
Year of publishing:	2008
Source:	Maribor
Typology:	2.09 - Master's Thesis
Organization:	UM FERI - Faculty of Electrical Engineering and Computer Science
Publisher:	[A. Zamuda]
UDC:	519.8:004.8
COBISS:	12461846
Views:	4045
Downloads:	360
Average score:	0 (0 votes)
Metadata:

Other data

Secondary language:	English
Secondary title:	Self-adaptation of control parameters in differential evolution for multiobjective optimization algorithm
Secondary abstract:	We present a new differential evolution algorithm for multiobjective optimization controlled by the self-adaptation mechanism introduced in evolution stategies. Algorithm design is presented with mathematically formal notation of algorithm's main parts and their assembly. The algorithm is described using a pseudocode. Computational complexity of the algorithm is given and some empirical measurement are given for evidence. Self-adaptation dynamics of control parameters is also studied. State of the art test problems and quality indicators from literature for performance assessment of multiobjective optimization algorithms are listed. Using these, performance assessments of the algorithm are obtained showing numerous statistically significant improvements. Obtained results with the algorithm are also compared with related algorithms and statistically significant differences of the compared algorithms are pointed out on empirical results.
Secondary keywords:	evolutionary algorithms;differential evolution;control parameters self-adaptation;multiobjective optimization;
URN:	URN:SI:UM:
Type (COBISS):	Master's thesis
Thesis comment:	Univ. v Mariboru, Fak. za elektrotehniko, računalništvo in informatiko
Pages:	XVI, 112 str.
Keywords (UDC):	mathematics;natural sciences;naravoslovne vede;matematika;mathematics;matematika;operational research (or): mathematical theories and methods;operacijsko raziskovanje;science and knowledge;organization;computer science;information;documentation;librarianship;institutions;publications;znanost in znanje;organizacije;informacije;dokumentacija;bibliotekarstvo;institucije;publikacije;prolegomena;fundamentals of knowledge and culture;propaedeutics;prolegomena;splošne osnove znanosti in kulture;computer science and technology;computing;data processing;računalniška znanost in tehnologija;računalništvo;obdelava podatkov;artificial intelligence;umetna inteligenca;
ID:	13726