Ohio State researchers and students gathered in Pomerene Hall on Sept. 24 for two events exploring transfer learning: a workshop featuring Ohio State faculty and graduate students, followed by a seminar with Yang Feng, professor of biostatistics at New York University’s School of Global Public Health.

Presented as part of TDAI’s yearlong research theme The Geometrization of AI, the morning workshop brought together researchers working on methods that help machine learning models apply knowledge across related tasks and domains. The presentations offered perspectives on transfer learning from across statistics, data science and AI research.

In the afternoon seminar, “Adaptive Knowledge Transfer,” Feng examined how researchers can identify useful data sources and draw on them when labeled data are limited or expensive to obtain. His presentation addressed representation multi-task learning and approaches to unsupervised and federated transfer learning, as well as challenges including differences between data sources, negative transfer, high dimensionality and data privacy.

Together, the events gave Ohio State’s research community an opportunity to examine both the statistical foundations of transfer learning and its potential for interdisciplinary AI applications.