Posted by mriekert on Tue 13 Feb 2018, 06:37:38 Tue 13 Feb 2018, 06:37:38

Date: Friday 16 Feb
Time: 14:00
Venue: Tswelopele
Title: Multi-Guided Particle Swarm Optimization: A Multi-objective Particle Swarm Optimizer

Abstract:
In this thesis, a novel visualization technique is presented and applied to perform an exploratory analysis of the Vector Evaluated Particle Swarm Optimization (VEPSO) algorithm in low dimensional objective space. The exploratory analysis together with a quantitative analysis revealed that the VEPSO algorithm continues to explore without exploiting the wellperforming areas of the search space. A detailed investigation into the influence the choice of archive implementation have, on the performance of the VEPSO algorithm, is presented. Both the solution diversity and convergence is considered during the investigation. To better compare the performance, of two multi-objective optimization (MOO) algorithms, the application of attainment surfaces is investigated. The newly developed porcupine measure is presented to objectively compare algorithms using attainment surfaces in multi-dimensional objective space. Loosely based on the VEPSO algorithm, the multi-guided particle swarm optimization (MGPSO) algorithm is presented and evaluated. The results indicate that the MGPSO algorithm overcomes the weaknesses of the VEPSO algorithm and also outperforms a number of state-of-the-art MOO algorithms on at least two benchmark problem sets. Techniques developed in this study may also be widely applied to various other artificial intelligence fields.