Hybrid Learning of RBF Networks

0404886 - UIVT-O 20020228 RIV CZ eng J - Journal Article
Neruda, Roman - Kudová, Petra
Hybrid Learning of RBF Networks.
Neural Network World. Roč. 12, č. 6 (2002), s. 573-585. ISSN 1210-0552
R&D Projects: GA ČR GA201/00/1489
Institutional research plan: AV0Z1030915
Keywords : neural networks * RBF networks * gradient algorithm * three step learning * genetic algorithm
Subject RIV: BA - General Mathematics

Three different learning methods for RBF networks and their combinations are presented. Their performance is compared on two benchmarks problems: Two spirals and Iris plants. The results show that the three-step learning is usually the fastest, while the gradient learning achieves better precision.
Permanent Link: http://hdl.handle.net/11104/0003463