English

$\ell_1$ Regression using Lewis Weights Preconditioning and Stochastic Gradient Descent

Data Structures and Algorithms 2018-06-04 v3

Abstract

We present preconditioned stochastic gradient descent (SGD) algorithms for the 1\ell_1 minimization problem minxAxb1\min_{x}\|A x - b\|_1 in the overdetermined case, where there are far more constraints than variables. Specifically, we have ARn×dA \in \mathbb{R}^{n \times d} for ndn \gg d. Commonly known as the Least Absolute Deviations problem, 1\ell_1 regression can be used to solve many important combinatorial problems, such as minimum cut and shortest path. SGD-based algorithms are appealing for their simplicity and practical efficiency. Our primary insight is that careful preprocessing can yield preconditioned matrices A~\tilde{A} with strong properties (besides good condition number and low-dimension) that allow for faster convergence of gradient descent. In particular, we precondition using Lewis weights to obtain an isotropic matrix with fewer rows and strong upper bounds on all row norms. We leverage these conditions to find a good initialization, which we use along with recent smoothing reductions and accelerated stochastic gradient descent algorithms to achieve ϵ\epsilon relative error in O~(nnz(A)+d2.5ϵ2)\tilde{O}(nnz(A) + d^{2.5} \epsilon^{-2}) time with high probability, where nnz(A)nnz(A) is the number of non-zeros in AA. This improves over the previous best result using gradient descent for 1\ell_1 regression. We also match the best known running times for interior point methods in several settings. Finally, we also show that if our original matrix AA is approximately isotropic and the row norms are approximately equal, we can give an algorithm that avoids using fast matrix multiplication and obtains a running time of O~(nnz(A)+sd1.5ϵ2+d2ϵ2)\tilde{O}(nnz(A) + s d^{1.5}\epsilon^{-2} + d^2\epsilon^{-2}), where ss is the maximum number of non-zeros in a row of AA. In this setting, we beat the best interior point methods for certain parameter regimes.

Keywords

Cite

@article{arxiv.1708.07821,
  title  = {$\ell_1$ Regression using Lewis Weights Preconditioning and Stochastic Gradient Descent},
  author = {David Durfee and Kevin A. Lai and Saurabh Sawlani},
  journal= {arXiv preprint arXiv:1708.07821},
  year   = {2018}
}

Comments

31 pages, COLT 2018

R2 v1 2026-06-22T21:23:51.148Z