Adaptive Newton Sketch: Linear-time Optimization with Quadratic
Convergence and Effective Hessian Dimensionality
Jonathan Lacotte 1 Yifei Wang 1 Mert Pilanci 1
Abstract 1. Introduction
We propose a randomized algorithm with We consider a composite optimization problem of the form
quadratic convergence rate for convex optimiza-
tion problems with a self-concordant, composite, x : = argmin {f (x) : = f0 (x) + ...


雷达卡




京公网安备 11010802022788号







