Enhanced machine learning aided RF-SOI EDMOS reliability prediction

Abstract

The reliability of Radio Frequency Silicon-On-Insulator (RF-SOI) Lateral Extended Drain Mental-Oxide-Semiconductor Field-effect Transistor (EDMOS) is crucial for ensuring the performance and longevity of high-performance electronic devices as they are the fundamental building blocks of those devices. In this project, I present a comprehensive evaluation of Long Short-Term Memory (LSTM) networks as an effective means to improve the prediction accuracy of RF-SOI EDMOS reliability. My research is built upon previous work that employed Artificial Neural Networks (ANN) for reliability prediction [1]. In my investigation, we compared the performance of LSTM networks with other machine learning models, including Simple Recurrent Neural Networks (RNN) and Convolutional Neural Networks (CNN). The LSTM model demonstrated a much more accurate prediction with the same data set provided.