Robust frequent directions with application in online learning

Luo Luo, Cheng Chen, Zhihua Zhang, Wu Jun Li, Tong Zhang

Research output: Contribution to journalArticlepeer-review

19 Scopus citations

Abstract

The frequent directions (FD) technique is a deterministic approach for online sketching that has many applications in machine learning. The conventional FD is a heuristic procedure that often outputs rank deficient matrices. To overcome the rank deficiency problem, we propose a new sketching strategy called robust frequent directions (RFD) by introducing a regularization term. RFD can be derived from an optimization problem. It updates the sketch matrix and the regularization term adaptively and jointly. RFD reduces the approximation error of FD without increasing the computational cost. We also apply RFD to online learning and propose an effective hyperparameter-free online Newton algorithm. We derive a regret bound for our online Newton algorithm based on RFD, which guarantees the robustness of the algorithm. The experimental studies demonstrate that the proposed method outperforms state-of-the-art second order online learning algorithms.

Original languageEnglish
JournalJournal of Machine Learning Research
Volume20
StatePublished - 1 Feb 2019
Externally publishedYes

Keywords

  • Frequent directions
  • Matrix approximation
  • Online Newton algorithm
  • Online convex optimization
  • Sketching

Fingerprint

Dive into the research topics of 'Robust frequent directions with application in online learning'. Together they form a unique fingerprint.

Cite this