Topology optimization of analog circuit design via global optimization using factorization machines with digital annealer

Abstract

This study delves into the automatic analog circuit design, focusing primarily on topology optimization. Many real-world calculations and evaluations often involve complex and nonlinear problems, posing significant challenges in optimization. Obtaining characteristic values using a simulator can require significant computational resources, adding to the complexity of these issues. In this paper, we employ an innovative optimization method that employs a factorization machine with a digital annealer (FM-DA) and a genetic algorithm (GA). The result is a global optimization method that integrates local optimization structures into its framework. FM-DA&GA represents a black-box optimization technique, where the input variables are expressed as binary variables. The optimization process is then performed using factorization machines that predict characteristic data. This method has been proven to be highly effective for topology optimization in the design of analog circuits. Furthermore, this method can produce a structure with excellent optimization performance at a less computational cost owing to the use of global optimization.