EVENTS

In recent years, the field of Machine Learning has made significant progress in theory and applications. This success is rooted in the mutual stimulation of mathematical insight and experimental studies. On the one hand, mathematics allows to conceptualize and formalize core problems within learning theory, leading, for instance, to performance bounds for learning algorithms. On the other hand, experimental studies confirm theoretical predictions and instigate new directions in theoretical research. This meeting aims to discuss the interaction between theory and practice, with focus on the current gaps between the two. The talks will be centered around themes including the following.

• Gradient Methods (gradient optimization, stochastic gradient, natural gradients, gradient applied to deep networks, ...),
• Natural Geometric Structures (Information Geometry, optimal transport geometry, ...),
• Generalisation Theory (statistical learning theory, complexity measures, Ill-posed inverse problems, regularization, implicit bias...)
• Functional analytical tools (approximation theory, harmonic analysis, ...)
• Overparametrization and random matrix theory (neural tangent kernel, lazy training, convergence of gradient descent, generalization bounds, ...)