An exact and direct analytical method for the design of optimally robust CNN templates

In this paper, we present an analytical design approach for the class of bi polar cellular neural networks (CNN's) which yields optimally robust templa te parameters. We give a rigorous definition of absolute and relative robus tness and show that all well-defined CNN tasks are characterized by a finit e set of linear and homogeneous inequalities. This system of inequalities c an be analytically solved for the most robust template by simple matrix alg ebra. For the relative robustness of a task, a theoretical upper bound exis ts and is easily derived, whereas the absolute robustness can be arbitraril y increased by template scaling. A series of examples demonstrates the simp licity and broad applicability of the proposed method.