Bridging Statistics and Modern AI: Foundations for Deep Learning and Generative Models

Deep learning and generative AI have achieved unprecedented practical success, yet their scale and complexity outpace classical statistical theory and call for new mathematical understanding. Fundamental questions remain about why these models generalize, how to efficiently adapt them in high-dimensional regimes, and what statistical principles underlie their generative and biologically inspired learning mechanisms. This session highlights recent advances that bring rigorous mathematical and statistical tools to bear on these challenges, offering theory-driven perspectives that demystify modern architectures, reveal structure in unstructured data, and establish principled foundations for scalable and interpretable learning. By bridging classical insights with contemporary AI practice, the session underscores the central role of mathematics and statistics in explaining and guiding the future of deep learning and generative modeling.