A controller to improve the convergence of a neural network

A controller that improves the performance of the Hopfield (feedback) model of neural networks is described. In the asynchronous Hopfield model, a random selection is made in each step of the interaction, whereas in a model with a controller, the choice of a neuron for the next update is more selective. The controller selects responses in a way that further helps to arrive at correct memory. An external input is applied to neurons in the net, increasing the capacity of the network and helping its performance in the sense that the final response to a given probe is usually closer in terms of Hamming distance to the probe. It is shown that the model with the controller has a somewhat higher capacity