Abstract
This presentation introduces a new evolutionary algorithm that can be used to find starting values (SVs)for nonlinear regression (NLR) problems. The method is called a Stochastic Funnel Algorithm (SFA). The SFA is introduced through its operation in the SSE/parameter space of a NLR model. The SFA uses a relaxed trust region in the log10 space of the parameters of an NLR model. The SFA takes explicit advantage of the general shape of the Sum-of-Squares Error (SSE) of a data/model set to explore and exploit that space and eventually converge to the region of the best parameter estimates.
Original language | English |
---|---|
Title of host publication | WORLDCOMP'08 |
Subtitle of host publication | World Congress in Computer Science, Computer Engineering and Applied Computing |
Editors | Hamid R Arabnia, Youngsong Mun |
Place of Publication | USA |
Publisher | CSREA Press |
Pages | 96-102 |
Number of pages | 7 |
ISBN (Electronic) | 1601320698 |
Publication status | Published - 2008 |
Event | International Conference on Genetic and Evolutionary Methods (GEM) - Las Vegas, Nevada, USA, New Zealand Duration: 14 Jul 2008 → 17 Jul 2008 |
Conference
Conference | International Conference on Genetic and Evolutionary Methods (GEM) |
---|---|
Country/Territory | New Zealand |
Period | 14/07/08 → 17/07/08 |