English

Doubly robust nearest neighbors in factor models

Machine Learning 2024-01-31 v3 Machine Learning

Abstract

We introduce and analyze an improved variant of nearest neighbors (NN) for estimation with missing data in latent factor models. We consider a matrix completion problem with missing data, where the (i,t)(i, t)-th entry, when observed, is given by its mean f(ui,vt)f(u_i, v_t) plus mean-zero noise for an unknown function ff and latent factors uiu_i and vtv_t. Prior NN strategies, like unit-unit NN, for estimating the mean f(ui,vt)f(u_i, v_t) relies on existence of other rows jj with ujuiu_j \approx u_i. Similarly, time-time NN strategy relies on existence of columns tt' with vtvtv_{t'} \approx v_t. These strategies provide poor performance respectively when similar rows or similar columns are not available. Our estimate is doubly robust to this deficit in two ways: (1) As long as there exist either good row or good column neighbors, our estimate provides a consistent estimate. (2) Furthermore, if both good row and good column neighbors exist, it provides a (near-)quadratic improvement in the non-asymptotic error and admits a significantly narrower asymptotic confidence interval when compared to both unit-unit or time-time NN.

Keywords

Cite

@article{arxiv.2211.14297,
  title  = {Doubly robust nearest neighbors in factor models},
  author = {Raaz Dwivedi and Katherine Tian and Sabina Tomkins and Predrag Klasnja and Susan Murphy and Devavrat Shah},
  journal= {arXiv preprint arXiv:2211.14297},
  year   = {2024}
}
R2 v1 2026-06-28T07:13:04.208Z