多分布学习的样本复杂度

Dec, 2023

The sample complexity of multi-distribution learning

Binghui Peng

TL;DR这篇论文研究多分布学习，给出了一个样本复杂度为$\widetilde{O}((d+k)\epsilon^{-2}) \cdot (k/\epsilon)^{o(1)}$的算法，解决了COLT 2023的开放问题。

Abstract

multi-distribution learning generalizes the classic pac learning to handle data coming from multiple distributions. Given a set of $k$ data distributions and a →