Exact design of reciprocal cellular neural networks

Based on two classes of equilibrium equations, a design method for reciprocal cellular neural networks is presented. The local rules defining the task to be accomplished by the network are directly mapped into a set of linear inequalities that bound the solution space of the network parameters for the given problem. All points in the solution space guarantee the correct operation of the network. A solution can be computed by the relaxation method for solving sets of linear inequalities.<>