Chengchao Zhao, Beijing Computational Science Research Center
Room 306, No.5 Science Building
We introduce Weinan E, Chao Ma and Lei Wu’s work: optimization and generalization are two central issues in the theoretical analysis of machine learning models. They show that gradient descent dynamics can achieve zero training loss exponentially fast regardless of the quality of the labels and give some results on optimization and generalization.
A Comparative Analysis of the Optimization and Generalization Property of Two-layer Neural Network and Random Feature Models Under Gradient Descent Dynamics