no code implementations • 20 Feb 2020 • Thomas Flynn, Kwang Min Yu, Abid Malik, Nicolas D'Imperio, Shinjae Yoo
This work examines the convergence of stochastic gradient-based optimization algorithms that use early stopping based on a validation function.
no code implementations • 25 Sep 2019 • Thomas Flynn, Kwang Min Yu, Abid Malik, Shinjae Yoo, Nicholas D'Imperio
This work examines the convergence of stochastic gradient algorithms that use early stopping based on a validation function, wherein optimization ends when the magnitude of a validation function gradient drops below a threshold.