Tutorial 2
Tom Dhane, Ghent University - imec
Yens Lindeman, Ghent University - imec
Abstract -
This lecture introduces the core principles of machine learning and macromodeling, with a particular focus on delayed electronic systems, such as high-speed interconnects and advanced electronic packaging structures. These systems exhibit complex frequency-dependent behavior and propagation delays, posing significant challenges for conventional modeling techniques. The talk highlights data-efficient learning strategies and physics-informed, kernel-based system-identification methods that explicitly account for these delays while providing accurate, uncertainty-aware models. Particular attention is given to the development of compact, delay-aware macromodels that capture the underlying system dynamics without requiring unnecessarily high model orders. By combining machine learning with established macromodeling techniques, these approaches enable efficient and reliable frequency-domain characterization of complex electronic systems, with applications in signal and power-integrity analysis.
Tom Dhaene
is a Full Professor at Ghent University (Belgium) at IDLab (Internet Technology and Data Science Lab) of the Faculty of Engineering and Architecture since October 2007. He is also affiliated with imec.
Earlier, he gained industrial experience (1993-2005) at Alphabit (an imec spin-off), Hewlett-Packard and Agilent Technologies (now part of Keysight Technologies), and held an academic position at the University of Antwerp (2000-2009).
His current research interests include data-efficient machine learning (DE-ML), surrogate modeling, electronic design automation, system identification, signal integrity, electromagnetic compatibility, and numerical analysis techniques.
Yens Lindeman Yens Lindemans received the M.S. degree in engineering physics in 2023, from Ghent University, Ghent, Belgium, where he is currently pursuing the Ph.D. degree in electrical engineering at the Internet Technology and Data Science Lab (IDLab). His research interests include the application of machine learning and surrogate modeling techniques for the efficient simulation, modeling, and optimization of complex high-frequency electronic systems.