Large AI Model Native Wireless Networks
this is a DLT academic events supported by IEEE comsoc and tongji niveristy.
Abstract: Large artificial intelligence model (LAIM) has been tremendously developed in terms of capabilities and applications in recent years. Meanwhile, 6G mobile technologies are in the process of fast development and standardization. 6G is widely believed to be AI-native. However, the application of LAIM in wireless optimization is still very limited, and optimized deployment of LAIM in wireless networks is still largely unknown. We will first present integration of LAIM with traditional optimization algorithms to solve complex design problems in Integrated Sensing and Communication (ISAC) networks. Then we will develop a framework of collaborative agent LLM in wireless networks. Our results show LAIM has great potentials for optimizing and design wireless networks. Then we will discuss the optimized deployment of LAIM in wireless edge networks. In particular, we will discuss a pruning-aware and quantization-based LAIM co-inference scheme, where a pre-trained LAIM is pruned or quantized and partitioned into on-device and on-server sub-models for deployment. We show that the LAIM output distortion is upper bounded by its parameter distortion. Then a lower bound on parameter distortion via the rate-distortion theory, analytically capturing the relationship between pruning ratio and co-inference performance is given. The analytic results enable the optimized deployment of LAIM in wireless networks.
Date and Time
Location
Hosts
Registration
-
Add Event to Calendar
- 4800 Cao'an Highway
- shanghai, Shanghai
- China 201804
- Building: Zhixin Building, Tongji University,
- Room Number: Room 703,
- Contact Event Host
-
(chaowang@tongji.edu.cn )
- Co-sponsored by Prof. Chao Wang
Speakers
Ming Xiao of KTH
Large AI Model Native Wireless Networks
Abstract: Large artificial intelligence model (LAIM) has been tremendously developed in terms of capabilities and applications in recent years. Meanwhile, 6G mobile technologies are in the process of fast development and standardization. 6G is widely believed to be AI-native. However, the application of LAIM in wireless optimization is still very limited, and optimized deployment of LAIM in wireless networks is still largely unknown. We will first present integration of LAIM with traditional optimization algorithms to solve complex design problems in Integrated Sensing and Communication (ISAC) networks. Then we will develop a framework of collaborative agent LLM in wireless networks. Our results show LAIM has great potentials for optimizing and design wireless networks. Then we will discuss the optimized deployment of LAIM in wireless edge networks. In particular, we will discuss a pruning-aware and quantization-based LAIM co-inference scheme, where a pre-trained LAIM is pruned or quantized and partitioned into on-device and on-server sub-models for deployment. We show that the LAIM output distortion is upper bounded by its parameter distortion. Then a lower bound on parameter distortion via the rate-distortion theory, analytically capturing the relationship between pruning ratio and co-inference performance is given. The analytic results enable the optimized deployment of LAIM in wireless networks.
Biography:
Biography: Ming Xiao received received Ph.D degree from Chalmers University of technology, Sweden in November 2007. From November 2007 to now, he has been in the Department of Information Science and Engineering, Royal Institute of Technology, KTH, Sweden, where he is currently a full Professor. He was an Editor for IEEE Transactions on Communications (2012-2017), and an Editor for IEEE Transactions on Wireless Communications from 2018 to 2025 and an area editor for IEEE Open Journal of the Communication Society from 2019 to 2024. He is presently an Area Editor for IEEE Transactions on Communications. He is also a TPC co-chair of IEEE International Conference in Communications, 2028. He received IEEE Vehicular Technology Society Best Magazine Paper Award 2023.
Email:
Address:Stockholm, Stockholms lan, Sweden