Variance Reduced Stochastic Gradient Descent with Neighbors

Thomas HofmannAurelien LucchiSimon Lacoste-JulienBrian McWilliams

Stochastic Gradient Descent (SGD) is a workhorse in machine learning, yet its slow convergence can be a computational bottleneck. Variance reduction techniques such as SAG, SVRG and SAGA have been proposed to overcome this weakness, achieving linear convergence... (read more)

PDF Abstract NeurIPS 2015 PDF NeurIPS 2015 Abstract