Autoassociative memory using refractory period of neurons and its on-line learning

In the present paper, we propose a novel autoassociative memory model of the neural network consisting of neurons which enter refractory period according to a threshold. We, furthermore, propose the refractory threshold made to change adaptively and autonomously based on network activity. The optimal network activity, then, is obtained by experiments on a static association model and the value is used to control the threshold. Finally, using network activity, the network with online learning mechanism is also proposed an it is shown that the network can detect unknown patterns and memorise them.