Gradient Descent with Identity Initialization Efficiently Learns Positive-Definite Linear Transformations by Deep Residual Networks

Peter L. Bartlett, David P. Helmbold, Philip M. Long. Gradient Descent with Identity Initialization Efficiently Learns Positive-Definite Linear Transformations by Deep Residual Networks. Neural Computation, 31(3), 2019. [doi]

Abstract

Abstract is missing.