Neironu tīkli stundu saraksta sastādīšanas modelī
Date
2007
Authors
Zuters, Jānis
Journal Title
Journal ISSN
Volume Title
Publisher
Latvijas Universitāte
Abstract
Stundu saraksta sastādīšanas problēma ir plaši sastopams grafiku (sarakstu) sastādīšanas problēmu veids. Šāda veida problēmas var tikt risinātas, izmantojot ģenētiskos algoritmus, tomēr ģenētiskie algoritmi paši par sevi veido tikai karkasu, kurā tiek ievietoti dažādi algoritmi, kas lielā daļā gadījumu ir specifiski risināmajai problēmai. Promocijas darba mērķis ir izstrādāt metodes, kas nodrošinātu neironu tīkla kā skaitļošanas sistēmas iesaistīšanu uz ģenētiskajiem algoritmiem balstītā stundu saraksta sastādīšanas modelī ar nolūku uzlabot stundu saraksta sastādīšanas procesā iegūto stundu sarakstu kvalitāti. Šim nolūkam tika veikta neironu tīklu izpēte, lai tos izmantotu kā derīguma funkcijas komponenti uz ģenētiskajiem algoritmiem bāzētā sistēmā, kas paredzēta skolu stundu sarakstu sastādīšanai.
Timetabling is a commonproblem of scheduling. Genetic algorithms (GAs) are considered to be a typical method for solving such problems, but GAs are used only to establish a framework into which various algorithms, mostly problem-specific ones, are inserted. The goal of the thesis is to propose a set of methods that are developed for the involvement of neural networks in a GA-based timetabling model so as to improve the quality of the timetables which are obtained. To design such methods, neural networks were investigated in order to see the effect of using them as a component of fitness function within a GA-based school timetabling system.
Timetabling is a commonproblem of scheduling. Genetic algorithms (GAs) are considered to be a typical method for solving such problems, but GAs are used only to establish a framework into which various algorithms, mostly problem-specific ones, are inserted. The goal of the thesis is to propose a set of methods that are developed for the involvement of neural networks in a GA-based timetabling model so as to improve the quality of the timetables which are obtained. To design such methods, neural networks were investigated in order to see the effect of using them as a component of fitness function within a GA-based school timetabling system.
Description
Elektroniskā versija nesatur pielikumus
Keywords
Datorzinātnes , Datorzinātne , Informācijas tehnoloģija, datortehnika, elektronika, telekomunikācijas, datorvadība un datorzinātne