×
Apr 3, 2024 · Abstract:We prove new convergence rates for a generalized version of stochastic Nesterov acceleration under interpolation conditions.
A combination of interpolation and growth conditions has been used to prove fast convergence rates for SGD with line-search [14] , with the stochastic Polyak ...
People also ask
Apr 3, 2024 · New convergence rates for a generalized version of stochastic Nesterov acceleration under interpolation conditions are proved and are ...
May 29, 2024 · Bibliographic details on Faster Convergence of Stochastic Accelerated Gradient Descent under Interpolation.
Apr 3, 2024 · We prove new convergence rates for a generalized version of stochastic Nesterov acceleration under interpolation conditions.
Preprints. Faster Convergence of Stochastic Accelerated Gradient Descent under Interpolation. A. Mishkin, M. Pilanci, M. Schmidt. 2024 [arXiv] ; Full Papers.
Under interpolation, we further show that all smooth loss functions with a finite-sum structure satisfy a weaker growth condition. Given this weaker condition, ...
▷ Under this growth condition, we prove that SGD converges as fast as full-batch gradient descent for (strongly)-convex and non-convex functions. ▷ We show that ...
These considerations demonstrate that (i) SGD shows rapid convergence in the interpolation regime and that (ii) data parallelism should be computationally ...