Symposium on Mathematical Foundations of Trustworthy Learning

In today's data-driven world, machine learning plays an increasingly pivotal role, but with this prominence comes the need for rigorous mathematical underpinnings to ensure trustworthiness. There has been a recent flurry of theoretical research activities that examine fundamental principles such as privacy, adversarial robustness, reproducibility, generalization, and other societal aspects of machine learning algorithms.  

Our aim is to create a forum where researchers at all career stages can engage in a vibrant exchange of ideas, foster collaborations, identify pressing challenges,  and collectively advance the mathematical foundations of trustworthy machine learning. This symposium will bring together a unique diverse mix of leading researchers from (i) different fields (such as theoretical computer science, statistics, optimization, causality), (ii) different parts of the world (in Europe, North America, and Asia), and (iii) a balanced group of junior and senior researchers. The planned session topics will include

• Theory of deep learning - stat. / generalization viewpoint
• Theory of deep learning - optimization viewpoint
• Privacy-aware machine learning
• Reproducible optimization and learning 
• Robustness under distribution shifts and adversarial attacks, causality
• Societal foundations of machine learning