Unsupervised learning of synaptic delays based on learning automata in an RBF-like network of spiking neurons for data clustering

In this paper, a new delay shift approach for learning in an RBF-like neural network structure of spiking neurons is introduced. The synaptic connections between the input and the RBF neurons are single delayed connections and the delays are adapted during an unsupervised learning process. Each synaptic connection in this network is modeled by a learning automaton. The action of the automaton associated with each connection is considered as the delay of the corresponding synaptic connection. It is shown through simulations that the clustering precision of the proposed network is considerably higher than that of the existing similar neural networks.