BriefGPT.xyz
Feb, 2018
无需条件数的参数化加速方法
Parametrized Accelerated Methods Free of Condition Number
HTML
PDF
Chaoyue Liu, Mikhail Belkin
TL;DR
提出一种参数化了的加速算法来应对条件数未知或误估的情况，并对几种重要的加速算法进行了谱层级分析，获得了显式表达式并提高了最坏情况下的收敛速率。
Abstract
Analyses of accelerated (momentum-based) gradient descent usually assume bounded
condition number
to obtain exponential
convergence rates
. However, in many real problems, e.g., kernel methods or deep neural netwo
→