AIBullisharXiv – CS AI · Mar 36/107
🧠
REMIND: Rethinking Medical High-Modality Learning under Missingness--A Long-Tailed Distribution Perspective
Researchers propose REMIND, a framework for medical multi-modal AI learning that addresses the challenge of missing data across multiple modalities. The solution uses a Mixture-of-Experts architecture to handle long-tail distributions of modality combinations and shows superior performance on real-world medical datasets.