BriefGPT.xyz
Nov, 2019
采样与优化之间的计算分离
Computational Separations between Sampling and Optimization
HTML
PDF
Kunal Talwar
TL;DR
该论文研究贝叶斯学习中常见的最大后验估计和后验分布采样的计算任务，证明在非凸情况下后验分布采样有时比优化更快，并展示两者在计算复杂性上的不可比较性，呈现出计算复杂度的急剧相变。
Abstract
Two commonly arising computational tasks in
bayesian learning
are
optimization
(Maximum A Posteriori estimation) and
sampling
(from the po
→