Parameter optimization of evolving spiking neural network with dynamic population particle swarm optimization
Evolving Spiking Neural Network (ESNN) is widely used in classification problem. However, ESNN like any other neural networks is incapable to find its own parameter optimum values, which are crucial for classification accuracy. Thus, in this study, ESNN is integrated with an improved Particle Swarm...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
2018
|
| Subjects: | |
| Online Access: | http://eprints.utm.my/id/eprint/81484/1/NurNadiahMdSaidMFC2018.pdf http://eprints.utm.my/id/eprint/81484/ http://dms.library.utm.my:8080/vital/access/manager/Repository/vital:119781 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Universiti Teknologi Malaysia |
| Language: | English |
| Summary: | Evolving Spiking Neural Network (ESNN) is widely used in classification problem. However, ESNN like any other neural networks is incapable to find its own parameter optimum values, which are crucial for classification accuracy. Thus, in this study, ESNN is integrated with an improved Particle Swarm Optimization (PSO) known as Dynamic Population Particle Swarm Optimization (DPPSO) to optimize the ESNN parameters: the modulation factor (Mod), similarity factor (Sim) and threshold factor (C). To find the optimum ESNN parameter value, DPPSO uses a dynamic population that removes the lowest particle value in every pre-defined iteration. The integration of ESNN-DPPSO facilitates the ESNN parameter optimization searching during the training stage. The performance analysis is measured by classification accuracy and is compared with the existing method. Five datasets gained from University of California Irvine (UCI) Machine Learning Repository are used for this study. The experimental result presents better accuracy compared to the existing technique and thus improves the ESNN method in optimising its parameter values. |
|---|