Dynamic recurrent neural networks for real time learning
Kiara de Deus Demura, M. Kajiura, Y. Anzai · 2005
The recurrent flash learning network is a real time learning architecture for dynamic systems. It does not use error-correction algorithms, and requires only a single presentation of training examples for learning of dynamics. The authors' method is simple and learns very quickly compared to the backpropagation through time (BPTT) learning algorithm.