18.11.2024: We are thrilled to share that our paper, “Machine Learning Based Data Validation for Signal Integrity and Power Integrity Using Supervised Time Series Classification”, has been accepted for publication in the IEEE Transactions on Electromagnetic Compatibility. This work is the result of a fruitful collaboration between researchers from the Institut für Theoretische Elektrotechnik of the Hamburg University of Technology and the Chair of Integrated Electronic Systems of the Otto-von-Guericke University Magdeburg Fabian Lurz.

In this work, we present a novel machine learning-based approach for validating data in signal integrity (SI) and power integrity (PI). By training time series classification networks on data labeled by expert engineers, we replicate human visual assessments to predict the agreement between two curves. In comparison to other error and similarity metrics, this method offers the advantage of capturing domain-specific expert knowledge in the comparison of curves. This approach allows the systematic, fast and objective validation of large datasets.

Details:
📌 Title: Machine Learning Based Data Validation for Signal Integrity and Power Integrity Using Supervised Time Series Classification.
📌 Authors: Youcef Hassab, Til Hillebrecht, Fabian Lurz, Christian Schuster.
📌 Journal: IEEE Transactions on Electromagnetic Compatibility.
📌 Date: October 2024.
📌 DOI: 10.1109/TEMC.2024.3450307.

30.10.2024: Excited to share our research “Efficient Iterative Data Generation Using Evaluation of Prioritized Input Parameters in ANNs for SAR Prediction in Human Head Models at 13.56 MHz”, by Hamideh Esmaeili, Cheng Yang and Prof. Christian Schuster. Our lates work has just been accepted for publication in IEEE Transactions on Electromagnetic Compatibility and is now available in IEEE Early Access!

In this work, using parameter prioritization combined with artificial neural networks (ANNs), we iteratively evaluate high-impact variables, focusing on mass-averaged SAR variations in human head models. We present an iterative dataset generation method that significantly reduces sample requirements by emphasizing key input parameters. By fine-tuning inputs with each iteration, our approach achieves >95% ML prediction accuracy with just 5–10 times the number of samples as the input parameters, resulting in a ~60% reduction in simulation time and memory use. This method effectively avoids redundancy and maintains accuracy, offering a scalable solution for high-dimensional BEM problems.
You can find more information here: https://doi.org/10.1109/TEMC.2024.3439468

18.09.2024: Large Signal Integrity PCB Dataset Created 
Excited to share our recent work, by Til Hillebrecht, Morten Schierholz, Youcef Hassab, Johannes Alfert and Christian Schuster, where we generated and analyzed a large dataset of PCB-based high-speed interconnect data for signal integrity (SI) using physics-based modeling and machine learning. This dataset, features frequencies up to 60 GHz and is available in the SI/PI-Database (https://lnkd.in/eRZquJmh). Our approach addresses the complexity of SI design, facilitates data-driven analysis in the SI domain. Find the article on IEEEXplore: https://lnkd.in/eDwKQRmM

31.07.2024: As part of the vacation program at Hamburg University of Technology our university recently hosted an exciting workshop using the BOB3 learning module to teach middle school students C++ in a fun, hands-on way! Together with Swantje Plambeck and Til Hillebrecht, the young coders engaged in an interactive sessions, solving problems, and bringing their little robot to life. Their enthusiasm and creativity were truly inspiring!