Redes neurais fuzzy aplicadas em identificação e controle de sistemas

TL;DRAbstract

This work compares the performance of neural fuzzy, neural network and fuzzy systems, to model and control non-linear dynamical systems. Due to the need of temporal representations, two recurrent neural fuzzy networks are proposed based on an hybrid static neural fuzzy architecture. Temporal processing is induced by local and global recurrence in the hidden layer neurons. A learning method based on gradient search and associative reinforcement learning is proposed. Computational experiments suggest that recurrent neural fuzzy networks provide an effective alternative to model and control non-linear dynamical systems.

Chat with Paper

AI Agents for this Paper