Optimality aspects of centers of gravity algorithm for statistical circuit design

The yield optimization problem is very hard to solve because it involves the maximization of an integral in a multidimensional space. To date, several yield optimizations based on Monte Carlo estimation have been proposed. The centers of gravity (CG) algorithm was originally developed on a heuristic basis. However, its implementation is very simple and it gives satisfactory results in practical problems. In the literature, a gradient-view interpretation and justification of the algorithm were given for Gaussian parameters distribution. In the same spirit, an extent of gradient justification of this method in the case of uniform parameters distribution is presented here