BriefGPT.xyz
Mar, 2019
在线非凸学习: 跟随扰动领袖是最优的
Online Non-Convex Learning: Following the Perturbed Leader is Optimal
HTML
PDF
Arun Sai Suggala, Praneeth Netrapalli
TL;DR
研究基于非凸损失的在线学习问题，证明了经典的Perturbed Leader算法在该设置下可达到最佳遗憾率，进一步证明乐观的FTPL算法在序列损失可预测时的遗憾界更优。
Abstract
We study the problem of
online learning
with
non-convex losses
, where the learner has access to an offline
optimization oracle
. We show th
→