[SEMINAR] NOVEMBER 9, 2023 – Kirill Glinskiy – Do Channel Models Matter?
We are happy to invite you to the next talk of Moscow Telecommunication Seminar which will be held at 17.00 (MSK, UTC+03:00), on Thursday, November 9, 2023, in Skype. Click here to show the seminar start time in your timezone.
Title: Evaluating ML-Based Channel Prediction Algorithms for URLLC in High-Mobility Scenarios: Do Channel Models Matter?
Speaker: Kirill Glinskiy, IITP RAS, HSE University
Abstract: This talk investigates the role of channel models in evaluating machine learning-based channel prediction algorithms for Ultra-Reliable Low Latency Communications (URLLC) in high-mobility scenarios. Notably, it uncovers how the choice of channel model can significantly impact algorithm performance. While simplified channel models tend to generate overly optimistic results, the use of more realistic models reveals performance challenges, at times even leading to worse outcomes than having no prediction, particularly at high speeds. Moreover, the relationship between prediction accuracy and model complexity is discovered to be significantly more complex.
Bio: Kirill Glinskiy is currently pursuing his PhD under the supervision of Evgeny Khorov. He received his MS degree with honors in applied mathematics and physics from the Moscow Institute of Physics and Technology in 2022. He is a researcher at the Wireless Networks Lab, Institute for Information Transmission Problems of the Russian Academy of Sciences. His research interests include the application of machine learning methods in wireless networks and testbeds.