Machine Learning Methods in High-Speed Channel Modeling




Dr. Tianjian Lu received the M.S. and Ph.D degrees in electrical engineering from the University of Illinois at Urbana-Champaign, in 2012 and 2016, respectively. Since 2016, he has been with Google Pixel Phone Team. He was a recipient of Best Student Paper Awards at ACES2015 and EDAPS2016. He was also a recipient of the P. D. Coleman Outstanding Research Award by the Department of Electrical and Computer Engineering, University of Illinois at Urbana-Champaign, in 2016.

The simulation techniques employed in characterizing high-speed channels for signal integrity can be computationally expensive. With machine learning methods, we take a different route and propose addressing the efficiency issue by taking advantage of the existing simulation data. The proposed approach requires no complex circuit simulations or substantial domain knowledge. The model training can be achieved within a reasonable amount of time over modern computing hardware, and the obtained models can be reused for future designs, which amortizes the training cost. Once the training concludes, prediction can be performed in a highly efficient manner.