LimNet-Flexible Learning Network Containing Linear Properties
Jinglu Hu, Kotaro Hirasawa, Kousuke Kumamaru · Journal of Advanced Computational Intelligence and Intelligent Informatics · 1999
We propose a flexible learning network of a class of linear models. A nonlinear black box system is transformed into a network of known and unknown nodes (node functions), where a linear model is introduced. Unknown nodes are parameterized using neurofuzzy models. The resulting learning network is interpreted as a linear model network (LimNet) and features useful linear properties and universal approximation ability.