Separation of timescales controls feature learning and overfitting in large neural networks
New Technologies in Mathematics Seminar Speaker: Pierfrancesco Urbani, Universite Paris-Saclay, CNRS, CEA, Institut de physique theorique Title: Separation of timescales controls feature learning and overfitting in large neural networks Abstract: To understand the inductive bias and generalization capabilities of large, overparameterized machine learning models, it is essential to analyze the dynamics of their training algorithms. […]