When Device Modeling Meets Machine Learning: Opportunities and Challenges (Invited)

Device modeling is essential for circuit simulations and designs in terms of constructing the circuit matrix equations of KCL and KVL. While there are classical methodologies, machine learning techniques are promising to bring innovations in the landscape of device modeling. This work reviews the device modeling from a top-down perspective, covering two different interpretations of modeling. Then the recent process in the domain-specific machine learning approaches is briefly summarized for logic and memory devices. The challenges ahead, for the machine learning model to support the industry’s practical needs, are analyzed. A concept of fusion model, by deeply merging device physics and neural networks, is also explained. CCS CONCEPTS • Hardware $\rightarrow$ Electronic design automation $\rightarrow$ Modeling and parameter extraction.

Paper

Full text

PDF

When Device Modeling Meets Machine Learning: Opportunities and Challenges (Invited)

Semantic Scholar · Computer Science · 2024

Abstract

Device modeling is essential for circuit simulations and designs in terms of constructing the circuit matrix equations of KCL and KVL. While there are classical methodologies, machine learning techniques are promising to bring innovations in the landscape of device modeling. This work reviews the device modeling from a top-down perspective, covering two different interpretations of modeling. Then the recent process in the domain-specific machine learning approaches is briefly summarized for logic and memory devices. The challenges ahead, for the machine learning model to support the industry’s practical needs, are analyzed. A concept of fusion model, by deeply merging device physics and neural networks, is also explained. CCS CONCEPTS

  • Hardware $\rightarrow$ Electronic design automation $\rightarrow$ Modeling and parameter extraction.

Similar papers

© 2026 NYSGPT2525 LLC