作者: Ireneusz Czarnowski , Piotr Jędrzejowicz
DOI: 10.1007/978-3-642-38658-9_4
关键词:
摘要: Radial Basis Function Neural Networks (RBFNs) are quite popular due to their ability discover and approximate complex nonlinear dependencies within the data under analysis. The performance of RBF network depends on numerous factors. One them is a value shape parameter. This parameter has direct impact transfer function each hidden unit. Values parameters, including its shape, set during RBFN tuning phase. Setting values can be viewed as optimization problem in which considered maximized. In paper agent-based population learning algorithm finding optimal or near proposed evaluated.