This review examines the use of artificial neural networks (ANNs) in the development of compact models for emerging devices. As new device technologies emerge, there is a critical need for compact models that are not only accurate and computationally efficient, but also sufficiently simple and physically consistent with support meaningful circuit-level simulation and design. We discuss how ANNs can be employed to address this challenge, emphasizing their potential to complement or extend traditional compact modeling approaches. The paper also outlines prospective research directions, focusing on the integration of machine learning techniques—particularly neural network architectures and training strategies—within the compact modeling framework. Finally, key methodological aspects are addressed, including the selection of training datasets, definition of model parameters, and relevant performance metrics, to guide the development of robust and interpretable models.
Cho et al. (Fri,) studied this question.