A Review on Hybrid Artificial Intelligence Systems for Integrating Circuit-Level Optimization with Machine Learning Algorithms

The rapid evolution of semiconductor technologies has fueled a significant surge in research on hybrid artificial intelligence (AI) systems, particularly for circuit-level optimization integrated with machine learning (ML). These systems address complex challenges in electronic design by leveraging multiple AI methodologies, such as neural networks, reinforcement learning, and statistical modeling. This review presents a comprehensive analysis of recent advancements in this domain, focusing on trends, methodologies, and global contributions. Between 2015 and 2024, research output grew exponentially, with publications rising from 2 in 2015 to 635 in 2024, underscoring increasing academic and industrial interest. Bibliometric analysis highlights leading sources, with fields like distributed computing and human-centered computing exhibiting the highest citation impacts. Prominent authors and influential publications reflect the intellectual foundations of this field. Countries like China, Saudi Arabia, and the United States dominate contributions, emphasizing global collaboration. Hybrid AI systems excel in optimizing analog and mixed-signal circuits through techniques such as Bayesian optimization, neural networks, and co-simulation with CAD/EDA tools, significantly enhancing efficiency and accuracy. Key applications include automated circuit sizing, fault diagnosis, and RF/microwave circuit modeling. Challenges such as data integration, computational overhead, and variability in manufacturing demand innovative solutions. Future research directions prioritize improving model interpretability, reducing computational costs, and integrating hybrid AI with emerging trends like IoT and edge computing. This review bridges theoretical innovations with practical implementations, contributing to the advancement of AI-driven integrated circuit design and optimization.