Learning dynamics and stability in networks with fuzzy synapses
K. Gopalsamy · 2006
ABSTRACT. A new model of Hopeld-type neural network of neurons with crisp somatic activa-tions which have some fuzzy synaptic modications is formulated which incorporates a Hebbian-type unsupervised learning algorithm. A set of sucient conditions are derived for the existence of a glob-ally exponentially stable steady state; the exponential convergence of the learning algorithm is also considered. Our model will reduce to one of fuzzy neural networks considered by others when the learning component is absent; when the fuzzy synapses are absent, our model will reduce to the well known Hopeld-type network AMS (MOS) Subject Classication. 92B20, 34K20 1.