95
4

Agnostic Sample Compression for Linear Regression

Abstract

We obtain the first positive results for bounded sample compression in the agnostic regression setting. We show that for p in {1,infinity}, agnostic linear regression with p\ell_p loss admits a bounded sample compression scheme. Specifically, we exhibit efficient sample compression schemes for agnostic linear regression in RdR^d of size d+1d+1 under the 1\ell_1 loss and size d+2d+2 under the \ell_\infty loss. We further show that for every other p\ell_p loss (1 < p < infinity), there does not exist an agnostic compression scheme of bounded size. This refines and generalizes a negative result of David, Moran, and Yehudayoff (2016) for the 2\ell_2 loss. We close by posing a general open question: for agnostic regression with 1\ell_1 loss, does every function class admit a compression scheme of size equal to its pseudo-dimension? This question generalizes Warmuth's classic sample compression conjecture for realizable-case classification (Warmuth, 2003).

View on arXiv
Comments on this paper