English

List Decodable Learning via Sum of Squares

Data Structures and Algorithms 2019-05-14 v1

Abstract

In the list-decodable learning setup, an overwhelming majority (say a 1β1-\beta-fraction) of the input data consists of outliers and the goal of an algorithm is to output a small list L\mathcal{L} of hypotheses such that one of them agrees with inliers. We develop a framework for list-decodable learning via the Sum-of-Squares SDP hierarchy and demonstrate it on two basic statistical estimation problems {\it Linear regression:} Suppose we are given labelled examples {(Xi,yi)}i[N]\{(X_i,y_i)\}_{i \in [N]} containing a subset SS of βN\beta N {\it inliers} {Xi}iS\{X_i \}_{i \in S} that are drawn i.i.d. from standard Gaussian distribution N(0,I)N(0,I) in Rd\mathbb{R}^d, where the corresponding labels yiy_i are well-approximated by a linear function \ell. We devise an algorithm that outputs a list L\mathcal{L} of linear functions such that there exists some ^L\hat{\ell} \in \mathcal{L} that is close to \ell. This yields the first algorithm for linear regression in a list-decodable setting. Our results hold for any distribution of examples whose concentration and anticoncentration can be certified by Sum-of-Squares proofs. {\it Mean Estimation:} Given data points {Xi}i[N]\{X_i\}_{i \in [N]} containing a subset SS of βN\beta N {\it inliers} {Xi}iS\{X_i \}_{i \in S} that are drawn i.i.d. from a Gaussian distribution N(μ,I)N(\mu,I) in Rd\mathbb{R}^d, we devise an algorithm that generates a list L\mathcal{L} of means such that there exists μ^L\hat{\mu} \in \mathcal{L} close to μ\mu. The recovery guarantees of the algorithm are analogous to the existing algorithms for the problem by Diakonikolas \etal and Kothari \etal. In an independent and concurrent work, Karmalkar \etal \cite{KlivansKS19} also obtain an algorithm for list-decodable linear regression using the Sum-of-Squares SDP hierarchy.

Keywords

Cite

@article{arxiv.1905.04660,
  title  = {List Decodable Learning via Sum of Squares},
  author = {Prasad Raghavendra and Morris Yau},
  journal= {arXiv preprint arXiv:1905.04660},
  year   = {2019}
}
R2 v1 2026-06-23T09:03:55.974Z