Design method for cellular neural network with linear relaxation

Based on the relaxation method for solving sets of linear inequalities, an algorithm for designing cellular neural networks (CNNs) has been developed. Equilibrium equations and initial conditions of the network are used to build subsets of linear inequalities. The symmetry conditions of templates are exploited as additional equality constraints. Using different initial conditions simultaneously, the authors are able to obtain more robust and reliable templates for a given problem. Simulation examples show that some robust templates, which are not sensitive to the initial conditions of the network, are generated by the application of the training rule. These templates may have an impact on the VLSI realization of CNNs