2019 (modified: 11 Nov 2022)ICML 2019Readers: Everyone
Abstract:Randomly initialized first-order optimization algorithms are the method of choice for solving many high-dimensional nonconvex problems in machine learning, yet general theoretical guarantees cannot...