AnalogGenie-Lite: Enhancing Scalability and Precision in Circuit Topology Discovery through Lightweight Graph Modeling

Jian Gao, Weidong Cao, Xuan Zhang

TL;DR: Lightweight graph modeling for enhancing circuit generative model's scalability and precision

Abstract:

The sustainable performance improvements of integrated circuits (ICs) drive the continuous advancement of nearly all transformative technologies. Since its invention, IC performance enhancements have been dominated by scaling the semiconductor technology. Yet, as Moore's law tapers off, a crucial question arises: ***How can we sustain IC performance in the post-Moore era?*** Creating new circuit topologies has emerged as a promising pathway to address this fundamental need. This work proposes AnalogGenie-Lite, a decoder-only transformer that discovers novel analog IC topologies with significantly enhanced scalability and precision via lightweight graph modeling.
AnalogGenie-Lite makes several unique contributions, including concise device-pin representations (i.e., advancing the best prior art from $O\left(n^2\right)$ to $O\left(n\right)$), frequent sub-graph mining, and optimal sequence modeling. Compared to state-of-the-art circuit topology discovery methods, it achieves $5.15\times$ to $71.11\times$ gains in scalability and 23.5\% to 33.6\% improvements in validity. Case studies on other domains' graphs are also provided to show the broader applicability of the proposed graph modeling approach. Source code: https://github.com/xz-group/AnalogGenie-Lite.

Lay Summary:

Integrated circuits (ICs) have powered decades of technological breakthroughs, enabling innovations from medical devices to quantum computing. Historically, IC improvements relied on making semiconductor components smaller, following Moore's Law. However, as physical limits are approached, simply shrinking components isn't enough.

AnalogGenie-Lite addresses this by discovering new analog circuit topologies—critical components that process continuous signals and bridge physical devices with digital systems. It uses a specialized AI model based on a simplified graph representation to accurately and efficiently explore new circuit designs. By reducing complexity, AnalogGenie-Lite achieves substantial improvements: it can handle larger circuits, generate valid designs with fewer errors, and discover novel designs that are unseen by humans.

This novel approach offers a practical path forward for maintaining performance advancements in electronics, even as traditional scaling methods reach their limits. Additionally, the techniques developed in AnalogGenie-Lite have potential applications beyond electronics, including areas like protein generation, personalized recommendation, and 3D object recognition.

Application-Driven Machine Learning: This submission is on Application-Driven Machine Learning.

Link To Code: https://github.com/xz-group/AnalogGenie-Lite

Primary Area: Applications->Chemistry, Physics, and Earth Sciences

Keywords: Graph modeling, Application of Generative Models, Electronic Design Automation

Submission Number: 4713