Conformal Inference for multivariate, complex, and heterogeneous data

Marcos Matabuena

University of Santiago de Compostela

In this project, in collaboration with Gábor Lugosi (UPF), we propose new uncertainty quantification methods based on the design of new Conformal Inference strategies for complex data that arise in modern personalized medicine applications. The new uncertainty methods can examine the reliability and safety of results obtained with machine learning in multiple medical analysis tasks. Furthermore, these methods constitute formal criteria to personalize and improve multiple medical decisions using different data-driven approaches. Our project constitutes a step towards automating medical decisions with machine learning techniques and potentially increasing efficiency in healthcare systems.