Computer Science

Dr Guoqiang Zhang

Senior Lecturer in Artificial Intelligence
Computer Science

Dr Guoqiang Zhang has been working as Senior Lecturer at Univerisity of Exeter.  His two main recent research activities are generative AI and distributed and parallel optimization. With regard to generative AI, he has been actively working on diffusion models. He focuses on how to solve the ordinary differential equations (ODEs) accurately to enable effective diffusion-based image/video sampling and editing. His recent research results have been published on top conferences like CVPR, ICLR, and ECCV. In addition to diffusion models, he has also been working on reversible transformer models for training large language models (LLMs). He has developed a new training method for transformers that saves training memory considerablly and improve generalization performance significantly.    

 
The second topic of distributed and parallel optimization intends to design effective message-passing algorithms to allow a number of nodes in a pear-to-pear network collaborate together in optimizing joint objective functions in an iterative manner. As the main investigator,  Guoqiang proposed a promising algorithm named primal-dual method of multipliers (PDMM) in 2018. Since then, he has been developing variants of PDMM with other researchers for different applications such as federated learning, and distributed training of deep learning models. The obtained research results are published on, for example, IEEE Trans. and ICML. 

 

See Google scholarhsip link for full publications 

View full profile