In 2025, a team from the University of Bristol achieved a significant accolade by winning an Outstanding Paper Award at the International Conference on Machine Learning (ICML), a leading global AI conference. Their paper, “Score Matching with Missing Data,” was among the six distinguished from 12,176 submissions for its exceptional quality. The research, conducted by PhD student Josh Givens, Associate Professor Song Liu, and Dr. Henry W. J. Reeve from Nanjing University, presents innovative methods for score matching in scenarios with incomplete data. Traditional score matching requires complete datasets, but the Bristol team introduced “marginal score matching,” a technique that imputes missing data during model training. This advancement holds promise for applications in areas like noisy image environments or genomics with degraded samples. Dr. Song Liu highlighted the award as a testament to Bristol’s prowess in foundational AI research, emphasizing its significance in tackling real-world data imperfections.
next post

