Early Stopping and Non-parametric Regression: An Optimal Data-dependent Stopping Rule

Provided by: Journal of Machine Learning Research (JMLR)
Topic: Big Data
Format: PDF
Early stopping is a form of regularization based on choosing when to stop running an iterative algorithm. Focusing on non-parametric regression in a reproducing kernel hilbert space, the authors analyze the early stopping strategy for a form of gradient-descent applied to the least-squares loss function. They propose a data-dependent stopping rule that does not involve hold-out or cross-validation data, and they prove upper bounds on the squared error of the resulting function estimate, measured in either the L2(P) and L2(Pn) norm.

Find By Topic