Qualitative analysis of electronic linear programming neural networks

One of the problems that can be solved by neural networks is the linear programming problem. Tank and Hopfield designed an electronic neural network and showed that the network is stable. A more general electronic neural network that solves the linear programming problem is presented herein and its error is analysed. A design with the lowest possible error is shown, detailed transient analysis of the network's outputs together with geometrical interpretations is performed. It is also shown how the network solutions depend on change of the network parameters. Some experimental examples are given.