Interconnection strengths and injected currents of a Hopfield neural net applied to an adaptive linear combiner

Mathematical expressions for the interconnection strengths and injected currents of a Hopfield type neural network are derived for the case where the network is expected to perform as an adaptive algorithm of an adaptive linear combiner (adaline). It is shown that the performance of the neural network is theoretically optimum. The problem is basically a quadratic programming one, and it is shown that the proposed neural network solution can be used for both cases where the variables are unconstrained and binary constrained. A representative application is given in the form of a simulation of a FIR filter.