Analog networks for solving convex nonlinear unconstrained programming problems without using gradient information of the objective function are proposed. The one-dimensional net can be used as a building block in multidimensional networks for optimizing objective functions of several variables.
