The title of Hakki’s award-winning paper is "Bayesian Active Learning for Uncertainty Quantification of High-Speed Channel Signaling.” His coauthors on the paper are his Ph.D. advisor, ECE Professor Madhavan Swaminathan, and Jose Hejase, Junyan Tang, and Dale Becker – all of IBM. 

Increasing data rates in server high-speed communication busses, such as in cloud data centers, makes their electrical performance more susceptible to uncertainties in manufacturing processes. As a result, it is essential to understand channel design limitations and performance under tolerances to ensure a robust system. However, predicting channel performance under tolerances can become very straining in time, effort, and computational resources. 

To address this issue, Torun and his colleagues present a new technique based on Bayesian Active Learning that automatically builds a predictive model from scratch, hence, does not require prior training data. While deriving the model, the presented algorithm simultaneously finds the worst case scenario to ensure channel compliancy and ranks the expected manufacturing tolerances in terms of their effect on electrical performance. The results show that the presented method provides a higher model accuracy in less CPU time compared to state-of-the-art methodologies and requires minimal human intervention.

Torun conducts his research in the Mixed Signal Design Group, located in the School of ECE. Swaminathan, who holds the John Pippin Chair in Microsystems Packaging & Electromagnetics, leads this lab.