Knowledge-Guided Neural Network based Nonlinear Current Model with Combined Loss Function
Junjun Qi, Hongliang Lü, Silu Yan, Zhiwu Jiang, Lin Cheng, Yuming Zhang · 2024
A knowledge-guided Neural network (KGNN)-based nonlinear current model of InP heterojunction bipolar transistor (HBT) is proposed. To improve modeling accuracy and consistency, a combined loss function is constructed and modeling consistency checks are implemented during the training of the neural network to penalize violations of the physical laws. The accuracy of the developed nonlinear current model was verified by comparing the measured and simulated pulse IV results.