Yi Liu's Homepage. 😀

Yi Liu (or Liu Yi，刘毅) is currently a first-year master’s student of Computer Science and Technology at the ADMIS Lab in the College of Computer Science and Technology of Tongji University, under the supervision of Associate Professor Wengen Li. Prior to this, he obtained his Bachelor’s degree in Computer Science and Technology from the College of Electronic and Information Engineering of Tongji University in 2024.

His research interests focus on various generative models, with a specialization in diffusion models. In 2024, he initiated his research career by delving into the low-level vision domain. During this period, he worked on improving the theoretical foundations of diffusion models for image restoration tasks, such as cloud removal in remote sensing image processing. Currently, he is dedicated to developing efficient generative models for diverse downstream applications (e.g., image restoration, image generation, etc.).

Yi Liu’s long-term objectives are twofold: to establish a rigorous mathematical understanding of generative models and to create hardware-aware optimizations for accelerating these models. Specifically, his goals encompass:

1. Understanding the mathematical principles underlying generative models (particularly diffusion models) through Bayesian statistics, differential equations, and optimal transport theory;
2. Accelerating the inference processes of generative models (e.g., enabling one-step sampling in diffusion models) by leveraging these theoretical insights;
3. Designing efficient network architectures (e.g., linear attention mechanisms, state space models) for generative models across different downstream tasks;
4. Optimizing the training and inference of generative models through hardware-aware programming using Triton, CUDA, and CuTe.

While these goals remain ongoing pursuits, Yi Liu is actively engaged in advancing this dual-focused research agenda through continuous study and practical implementation.

📧 If you are interested in Yi Liu’s work, please contact him via liuyi61@tongji.edu.cn or liuyi2052697@foxmail.com (preferred).