Conference Featured
2024 ConSinger: Efficient High-Fidelity Singing Voice Generation with Minimal Steps
We present a novel approach to solving X problem in machine learning...
Machine Learning
Probabilistic Models
Diffusion Models
Guorui Sang
PhD Student in Computer Science
Department of Computer Science • University of Illinois at Chicago
Machine LearningProbabilistic ModelsDiffusion Models
I am a PhD student in Computer Science with a focus on generative models, especially probabilistic methods like diffusion models
Research Interests
Publications
Recent research contributions and publications
2024
2023
Book
2023 《计算机考研精炼 1000 题》（Computer Science Graduate Examination: 1000 Exercises）
Published under the pen name '宇航'
Featured Projects
Open-source projects and research implementations
Research on Robotic Arm Simulation Using Diffusion Probabilistic Fields
FeaturedAug 2025 - Present
- Applied diffusion probabilistic fields to the generation of robotic arm trajectories.
- Created energy terms in the sampling stage, and used Langevin dynamics to steer the generated trajectory towards physically plausible quaternions and velocities, further utilized a Hamiltonian Neural Network to guide the torques applied to each joint to ground-truth torques.
- Demonstrated possibility of accelerating robotic simulation through GPU-based generative models and the ability to flexibly extend the length of the generated trajectory.
Research on Singing Voice Generation
FeaturedJul 2024 - Dec 2024
- Developed ConSinger, a consistency model-based singing voice synthesis system.
- Designed a novel scorer mechanism to identify optimal restoration points during training for efficient mel-spectrogram generation from music scores.
- Improved the generation quality (+1.16 MOS) at a 10% reduction in the generation speed.
Research on Pose Transfer Using Diffusion Models
FeaturedOct 2023 - Feb 2025
- Designed a three-branch U-Net structure that incorporates conditional image features effectively.
- Created a self-designed mixed classifier-free guidance to stably control the strength of different conditions.
- Demonstrated faster convergence, reduced 10%-50% parameters than SOTA models back then, and achieved comparable SSIM and PSNR.
Experience
Teaching Assistant
University of Illinois at Chicago
Chicago, IL
- Leading labs for Matlab course for 30+ students
- Held weekly office hours and graded exams and projects
- Developed supplementary course materials
Education
Ph.D. in Computer Science
University of Illinois at Chicago
Chicago, IL, USA
Advisor: Professor Pedram Rooshenas
Focus: Generative models, Probabilistic Models, Diffusion Models
Conducting research on robust and interpretable machine learning systems.
M.S. in Computer Science
Beijing University of Technology
Beijing, China
Specialized in conditional image generation using diffusion models.
B.S. in Software Engineering
Hainan University
Haikou, Hainan, China
Mastered essential knowledge in computer science and software engineering.
Skills & Expertise
Programming Languages
PythonC/C++CSS/HTML
Machine Learning
PyTorchPyTorch-Lignthing
Tools & Platforms
GitLinuxJupyter
Areas of Expertise
Machine LearningDeep LearningComputer VisionGenerative ModelsDiffusion Models