|Title:||Mutation particle swarm optimization for multilayer perceptron training with applications|
Particles (Nuclear physics)
Hong Kong Polytechnic University -- Dissertations
|Department:||Department of Electronic and Information Engineering|
|Pages:||vii, 42 leaves : ill. (some col.) ; 30 cm.|
|Abstract:||Particle Swarm Optimization (PSO) is one of the important evolutionary algorithms. It has been successfully applied to various optimization problems in science and engineering community. However, the traditional PSO suffers from the premature convergence problem. That is, the PSO easily falls into local minima rather than the global minima. In view of this, a new PSO, named Mutation PSO (MPSO), is proposed in this dissertation. Compared to the traditional PSO, there are two main differences in MPSO. Firstly, MPSO makes use of a mutation operator to update particles/individuals. The mutation operator was originally designed for Genetic Algorithm (GA), another widely used evolutionary algorithm. The use of mutation can increase the diversity of particles, as a result of which the premature convergence phenomenon can be alleviated. Secondly, a weighted update rule is proposed for MPSO to produce the new swarm. The traditional update rule used in PSO updates a particle's position based on its own best-known position and the best-known position in the swarm. By contrast, the proposed weighted update rule updates a particle's position based on not only its own best-known position and the best-known position in the swarm, but also a number of other good positions found in the current generation. These selected good positions are weighted according to their fitness values to update the swarm. The weighted update rule makes use of more information, and is expected to generate more effective particles for the next generation. The proposed MPSO is employed to train Multilayer Perceptron (MLP). Two tasks: namely curve fitting and classification, are used to evaluate the performance of the MLP trained by MPSO. In particular, the performance investigation is concentrated on scene classification. For a comparison purpose, MLPs trained using the error Backpropagation (BP), traditional PSO and GA are also investigated. The advantages and disadvantages of these algorithms are also analyzed. Experimental results demonstrate that the proposed MPSO outperforms PSO, BP and GA for the training of an MLP. In the last section, I conclude the dissertation with final remarks, and point out possible future work.|
|Rights:||All rights reserved|
Files in This Item:
|b24710623.pdf||For All Users (off-campus access for PolyU Staff & Students only)||578.6 kB||Adobe PDF||View/Open|
As a bona fide Library user, I declare that:
- I will abide by the rules and legal ordinances governing copyright regarding the use of the Database.
- I will use the Database for the purpose of my research or private study only and not for circulation or further reproduction or any other purpose.
- I agree to indemnify and hold the University harmless from and against any loss, damage, cost, liability or expenses arising from copyright infringement or unauthorized usage.
By downloading any item(s) listed above, you acknowledge that you have read and understood the copyright undertaking as stated above, and agree to be bound by all of its terms.
Please use this identifier to cite or link to this item: