Contrastive Representation Learning
Published:
Learning robust representations is fundamental to AI. Contrastive learning enables models to understand data by learning what is similar and what is different. My work applies these principles to medical imaging, with a focus on overcoming the challenges of limited labeled data and distribution shifts across imaging sites.
This research area includes self-supervised learning, unbiased contrastive frameworks, and multi-site harmonization techniques that improve generalization and clinical applicability.