Open-World Learning for Biomedical Data

This lecture discusses open-world learning for biomedical data, focusing on methods to leverage small datasets under different conditions. It covers the significance of single-cell multimodal omics, spatially resolved transcriptomics, and the challenges in generalizing across experiments. The instructor presents the development of novel machine learning paradigms for biomedical research, emphasizing the importance of discovering unknown phenomena and cell types. The lecture showcases the application of these methods in annotating spatially resolved single-cell data and the Human BioMolecular Atlas Program. It concludes with future directions in machine learning for biomedicine, including knowledge-grounded ML systems and multi-modal ML methods.