Optimizing Analog Circuit Design Through a Machine Learning‐Assisted Evolutionary Algorithm

ABSTRACT Evolutionary algorithms (EAs) based on circuit simulation are widely employed for analogue circuit sizing because of their high accuracy and adaptability in various cases. However, most of the existing research is focused on a limited set of analogue integrated circuit design specifications. When addressing a complete specification set, the extensive number of simulations required becomes impractical for large circuits. Recent studies incorporating machine learning (ML) techniques have accelerated the optimization process but still involve high simulation costs. This paper proposes an improved and efficient ML‐assisted evolutionary algorithm for analogue circuit sizing. The proposed approach integrates a machine learning model into the EA optimization process, effectively reducing the number of required simulations and improving the convergence speed. The experimental results demonstrate the efficiency of the proposed methodology in achieving reliable optimization, with a significant reduction in simulation cost and improved convergence.