Title of Talk: Multi-modal Semantic Communications for Multiple Tasks
Abstract: In the past decades, communications primarily focus on how to accurately and effectively transmit symbols. With the developments of cellular communication systems, the system capacity is gradually approaching to the Shannon limit. Inspired by the powerful deep learning technologies, semantic communications have been regarded as a promising solution to further improve the transmission efficiency, which aim to realize the successful semantic information exchange. In this talk, we will span from semantic communications for single modal data transmission with the single task to multimodal semantic communications for multi-tasks. Afterwards, the corresponding semantic-aware resource allocation optimization will be presented.
2:00-2:30pm Deniz Gunduz
Title: Generative Joint Source-Channel Coding
Abstract：In recent years, there have been significant developments in joint source-channel coding thanks to the deployment of deep neural networks (DNNs). The impressive deep joint source-channel coding (DeepJSCC) approach can deliver complex information sources, such as images and videos, over wideband wireless channels, taking care of compression, channel coding, modulation, and power allocation at once, while providing robustness against channel variations and channel estimation errors. In this talk, I will show that DeepJSCC can also benefit from powerful pretrained generative models to further improve the quality of reconstructed images. Generative DeepJSCC can deliver highly realistic images with high fidelity at even very low bandwidth and low signal-to-noise-ratio regimes, opening up new potentials for wireless gaming and metaverse applications.
Title: Electromagnetic Information Theory for MIMO Antenna Design
Abstract: This talk will introduce the electromagnetic (EM) information theory for MIMO antenna design. Firstly, I will overview the urgent requirement of future 5.5/6G. Secondly, I will show how EM theory governs the capacity limit of a MIMO communication system. Thirdly, I will connect the degree of freedom (DoF) in communication theory to the directivity in antenna theory. Finally, I will revisit the active element pattern and embedded efficiency of an antenna array and discuss their roles in affecting the capacity of the MIMO system. I will argue that the fundamental limits or physical bounds in antenna theory are essential to future MIMO communications including ultra-massive MIMO, holographic MIMO, etc.
Paper presentation3:00-3:30pm （Each has 5 minutes）:
- Downlink HMIMO Beamforming for Maximizing the Weighted Geometric Mean of SE
- Lightweight Generative Joint Source-Channel Coding for Semantic Image Transmission with Compressed Conditional GANs
- A Cooperative Relay Network for Semantic Communication with Network Coding
- Robust Federated Learning For Image Semantic Transmission Under Byzantine Attacks
- Semantic Information Extraction for Text Data with Probability Graph
- Semantic Integrity Measurement of Industrial Control Embedded Devices Based on National Secret Algorithm