English

Batch List-Decodable Linear Regression via Higher Moments

Machine Learning 2025-03-14 v1 Data Structures and Algorithms Statistics Theory Machine Learning Statistics Theory

Abstract

We study the task of list-decodable linear regression using batches. A batch is called clean if it consists of i.i.d. samples from an unknown linear regression distribution. For a parameter α(0,1/2)\alpha \in (0, 1/2), an unknown α\alpha-fraction of the batches are clean and no assumptions are made on the remaining ones. The goal is to output a small list of vectors at least one of which is close to the true regressor vector in 2\ell_2-norm. [DJKS23] gave an efficient algorithm, under natural distributional assumptions, with the following guarantee. Assuming that the batch size nn satisfies nΩ~(α1)n \geq \tilde{\Omega}(\alpha^{-1}) and the number of batches is m=poly(d,n,1/α)m = \mathrm{poly}(d, n, 1/\alpha), their algorithm runs in polynomial time and outputs a list of O(1/α2)O(1/\alpha^2) vectors at least one of which is O~(α1/2/n)\tilde{O}(\alpha^{-1/2}/\sqrt{n}) close to the target regressor. Here we design a new polynomial time algorithm with significantly stronger guarantees under the assumption that the low-degree moments of the covariates distribution are Sum-of-Squares (SoS) certifiably bounded. Specifically, for any constant δ>0\delta>0, as long as the batch size is nΩδ(αδ)n \geq \Omega_{\delta}(\alpha^{-\delta}) and the degree-Θ(1/δ)\Theta(1/\delta) moments of the covariates are SoS certifiably bounded, our algorithm uses m=poly((dn)1/δ,1/α)m = \mathrm{poly}((dn)^{1/\delta}, 1/\alpha) batches, runs in polynomial-time, and outputs an O(1/α)O(1/\alpha)-sized list of vectors one of which is O(αδ/2/n)O(\alpha^{-\delta/2}/\sqrt{n}) close to the target. That is, our algorithm achieves substantially smaller minimum batch size and final error, while achieving the optimal list size. Our approach uses higher-order moment information by carefully combining the SoS paradigm interleaved with an iterative method and a novel list pruning procedure. In the process, we give an SoS proof of the Marcinkiewicz-Zygmund inequality that may be of broader applicability.

Keywords

Cite

@article{arxiv.2503.09802,
  title  = {Batch List-Decodable Linear Regression via Higher Moments},
  author = {Ilias Diakonikolas and Daniel M. Kane and Sushrut Karmalkar and Sihan Liu and Thanasis Pittas},
  journal= {arXiv preprint arXiv:2503.09802},
  year   = {2025}
}
R2 v1 2026-06-28T22:18:12.756Z