Analyzing Optimization and Generalization in Deep Learning via Dynamics of Gradient Descent

This course focuses on analyzing optimization and generalization in deep learning through the dynamics of gradient descent. The learning outcomes include understanding optimization and generalization in classical machine learning and deep learning, as well as exploring gradient flow and end-to-end dynamics. The course teaches skills such as implicit preconditioning, matrix completion, deep matrix factorization, and analyzing the dynamics of singular values. The teaching method involves theoretical discussions and practical experiments. This course is intended for individuals interested in deep learning, optimization, and generalization in machine learning.