Abstract
Previously, weighted kernel regression (WKR) for solving small sample problems has been reported. The proposed WKR has been successfully employed to solve rational functions with very few samples. The design and development of WKR is important in order to extend the capability of the technique with various learning techniques. Based on WKR, a simple iteration technique is employed to estimate the weight parameters before WKR can be used in predicting the unseen test samples. In this paper, we investigate two learning techniques in estimating the weight parameters. For this purpose, a Ridge Regression (RR) and a guided search based on Particle Swarm Optimization (PSO) are used to investigate the capability of WKR in solving small sample problems. It is found that RR and PSO are better than iteration technique in terms of computational time and exibility of defining the objective function to estimate weight parameters, respectively, without sacrificing the quality of prediction, as supported by the conducted experiments.
Original language | English |
---|---|
Pages (from-to) | 705-710 |
Number of pages | 6 |
Journal | ICIC Express Letters |
Volume | 6 |
Issue number | 3 |
Publication status | Published - Mar 2012 |
Externally published | Yes |
Keywords
- Particle swarm optimization (PSO)
- Ridge regression (RR)
- Small samples
- Weighted kernel regression (WKR)
ASJC Scopus subject areas
- Control and Systems Engineering
- General Computer Science