Papers
Topics
Authors
Recent
Search
2000 character limit reached

An homotopy method for â„“p\ell_p regression provably beyond self-concordance and in input-sparsity time

Published 3 Nov 2017 in math.OC and cs.DS | (1711.01328v2)

Abstract: We consider the problem of linear regression where the ℓ2n\ell_2^n norm loss (i.e., the usual least squares loss) is replaced by the ℓpn\ell_p^n norm. We show how to solve such problems up to machine precision in O∗(n∣1/2−1/p∣)O^*(n^{|1/2 - 1/p|}) (dense) matrix-vector products and O∗(1)O^*(1) matrix inversions, or alternatively in O∗(n∣1/2−1/p∣)O^*(n^{|1/2 - 1/p|}) calls to a (sparse) linear system solver. This improves the state of the art for any p∉{1,2,+∞}p\not\in \{1,2,+\infty\}. Furthermore we also propose a randomized algorithm solving such problems in {\em input sparsity time}, i.e., O∗(Z+poly(d))O^*(Z + \mathrm{poly}(d)) where ZZ is the size of the input and dd is the number of variables. Such a result was only known for p=2p=2. Finally we prove that these results lie outside the scope of the Nesterov-Nemirovski's theory of interior point methods by showing that any symmetric self-concordant barrier on the ℓpn\ell_p^n unit ball has self-concordance parameter Ω~(n)\tilde{\Omega}(n).

Citations (53)

Summary

No one has generated a summary of this paper yet.

Paper to Video (Beta)

No one has generated a video about this paper yet.

Whiteboard

No one has generated a whiteboard explanation for this paper yet.

Open Problems

We haven't generated a list of open problems mentioned in this paper yet.

Continue Learning

We haven't generated follow-up questions for this paper yet.