|
Neuro-fuzzy systems are powerful hybridized systems withinthe
domains of arti-ficial neural networks (ANNs) and fuzzy
inference systems. Identifying a neur-o-fuzzy system is a
relevant issue having received extensive attention. Model-ing a
neuro-fuzzy system requires two steps: structure identification
and par-ameter identification. The former identifies the rough
structure of a system, and the latter fine-tunes detail
parameters of the system. Conventional appro-aches to identify a
system have their constraints. In order to overcome those
limitations, several intelligence systems have been applied in
this thesis. Inthis thesis, we propose an evolutionary model
that fulfills the two phases id-entifying a system:
simultaneoulsy identifying the structure and the paramete-rs.
The proposed model facilitates the construction of a neuro-fuzzy
system. This model provides an evolutionary approach to modify
a neuro-fuzzy system o-ther than conventional ones.
|