We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:


Current browse context:


Change to browse by:

References & Citations

DBLP - CS Bibliography


(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Machine Learning

Title: On Fast Convergence of Proximal Algorithms for SQRT-Lasso Optimization: Don't Worry About Its Nonsmooth Loss Function

Abstract: Many machine learning techniques sacrifice convenient computational structures to gain estimation robustness and modeling flexibility. However, by exploring the modeling structures, we find these "sacrifices" do not always require more computational efforts. To shed light on such a "free-lunch" phenomenon, we study the square-root-Lasso (SQRT-Lasso) type regression problem. Specifically, we show that the nonsmooth loss functions of SQRT-Lasso type regression ease tuning effort and gain adaptivity to inhomogeneous noise, but is not necessarily more challenging than Lasso in computation. We can directly apply proximal algorithms (e.g. proximal gradient descent, proximal Newton, and proximal Quasi-Newton algorithms) without worrying the nonsmoothness of the loss function. Theoretically, we prove that the proximal algorithms combined with the pathwise optimization scheme enjoy fast convergence guarantees with high probability. Numerical results are provided to support our theory.
Subjects: Machine Learning (cs.LG); Optimization and Control (math.OC); Machine Learning (stat.ML)
Cite as: arXiv:1605.07950 [cs.LG]
  (or arXiv:1605.07950v6 [cs.LG] for this version)

Submission history

From: Haoming Jiang [view email]
[v1] Wed, 25 May 2016 16:08:08 GMT (1263kb,D)
[v2] Mon, 11 Jul 2016 02:10:41 GMT (1263kb,D)
[v3] Mon, 2 Jan 2017 18:32:20 GMT (1148kb,D)
[v4] Sat, 3 Feb 2018 05:50:46 GMT (1242kb,D)
[v5] Wed, 14 Feb 2018 16:06:58 GMT (1194kb,D)
[v6] Sun, 14 Apr 2019 01:00:57 GMT (4177kb,D)

Link back to: arXiv, form interface, contact.