暂无评分数据
ICLR 2024
Locally Optimal Descent for Adaptive Stepsize Scheduling
TL;DR
We introduce a dynamic stepsize scheduler based on the local optimal step size, it attains optimal rate in the smooth (non-convex) setting and empirically shown to perform on par with manually fine-tuned schedulers.
摘要
关键词
step-size tuningoptimizationnon-convexlearning rate schedulers
评审与讨论
暂无评审记录