|国家预印本平台
首页|Submodular Functions Are Noise Stable

Submodular Functions Are Noise Stable

Submodular Functions Are Noise Stable

来源:Arxiv_logoArxiv
英文摘要

We show that all non-negative submodular functions have high {\em noise-stability}. As a consequence, we obtain a polynomial-time learning algorithm for this class with respect to any product distribution on $\{-1,1\}^n$ (for any constant accuracy parameter $\epsilon$). Our algorithm also succeeds in the agnostic setting. Previous work on learning submodular functions required either query access or strong assumptions about the types of submodular functions to be learned (and did not hold in the agnostic setting).

Homin K. Lee、Adam Klivans、Mahdi Cheraghchi、Pravesh Kothari

数学

Homin K. Lee,Adam Klivans,Mahdi Cheraghchi,Pravesh Kothari.Submodular Functions Are Noise Stable[EB/OL].(2011-06-02)[2025-08-05].https://arxiv.org/abs/1106.0518.点此复制

评论