RESEARCH

Semi-supervised Learning & Clustering

Semi-supervised learning addresses scenarios where labeled data is scarce but unlabeled data is abundant. My research leverages Optimal Transport to propagate labels from a small labeled set to the larger unlabeled portion of the data.

By computing transport plans between labeled and unlabeled distributions, supervision signals can be transferred in a geometrically meaningful way. I have also explored multi-view clustering through Optimal Transport, where complementary representations of the same data are aligned to reveal more robust cluster structures.