Attention-Based Machine Learning for Modeling of Non-Linear Circuits

Attention mechanism is a key part in present-day LLM. This talk explores the attention mechanism for modeling of high-speed/high-frequency non-linear circuits. The talk starts with an introduction of machine learning for non-linear microwave devices and circuits, highlighting the need for modeling short-term and long term memory effects, as well as the temporal-dynamical nature of non-linear modeling. The talk then describes attention mechanism and its relevance to the needs of non-linear circuit modeling, followed by a description of an attention-based deep recurrent neural network method for modeling non-linear circuits in high-speed/high-frequency electronic circuits and packages.