Introducing Transfer Learning Framework on Device Modeling by Machine Learning

In this study, we propose a novel transistor modeling method using machine learning techniques, with a focus on extrapolation performance. Our method leverages knowledge from a base model that is related to the target model, instead of relying solely on device-specific information. The results show that our approach outperforms other transistor modeling methods based on machine learning, particularly in modeling similar but different transistors that belong to the same device family. Our method was able to reduce the root mean squared error (RMSE) by up to 80.0% compared to other methods.