Accelerators and convergence measures for Monte-Carlo synthesis techniques

Abstract

Monte-Carlo synthesis techniques can be used to design new and complex systems that best meet a certain objective function with relative ease. Monte-Carlo synthesis is inefficient and does not provide obvious convergence measures. Accelerators based on probability distribution function shading and discriminant vector analysis are proposed. Convergence measures based on cluster identification and a statistical criterion are proposed. These enhancements are shown to significantly improve the performance of Monte-Carlo synthesis techniques. The implementation of these enhancements is shown through an example.