Abstract

Abstract: AI has demonstrated unprecedented performance in various application domains, including robotics, image processing, and language processing. We aim to showcase its application to modern communication systems. Specifically, we employ a transformer-based foundation model and multimodal time series representing communication system data to solve various downstream tasks. We develop methodologies to address challenges such as tokenization, positional embedding, multimodality, features of varying sizes, and normalization. Our results show comparable performance with benchmarks on all tasks, including CSI feedback, Doppler spectrum, and delay spread estimates.

Speakers:

Speaker 1:

Sattar Vakili is a Principal AI Research manager at MediaTek Research, the research arm of MediaTek — a globally renowned semiconductor company. He specializes in problems involving sequential decision-making in uncertain environments, with a focus on optimization, reinforcement learning, kernel-based modeling, and neural networks. Before joining MediaTek Research, Sattar worked at Second mind, a research lab in Cambridge, led by Professor Carl Rasmussen, Cambridge University. There, he gained expertise in kernel-based and Gaussian process models. Prior to that, he was a postdoc at Princeton University. Sattar earned his PhD at Cornell University in 2017 with a dissertation on sequential methods for learning and optimization.

Speaker 2:

Masoud Attarifar is a senior wireless software engineer at MediaTek. His expertise lies in wireless communication, communication theory, and signal processing. Prior to joining MediaTek, he was a postdoctoral researcher at UPF University under the supervision of Professor Angel Lozano. Masoud earned his PhD from Tehran University in 2019 with a dissertation on cell-free massive MIMO networks.