English

Stochastic Approximation for Canonical Correlation Analysis

Machine Learning 2018-02-27 v2 Machine Learning

Abstract

We propose novel first-order stochastic approximation algorithms for canonical correlation analysis (CCA). Algorithms presented are instances of inexact matrix stochastic gradient (MSG) and inexact matrix exponentiated gradient (MEG), and achieve ϵ\epsilon-suboptimality in the population objective in poly(1ϵ)\operatorname{poly}(\frac{1}{\epsilon}) iterations. We also consider practical variants of the proposed algorithms and compare them with other methods for CCA both theoretically and empirically.

Keywords

Cite

@article{arxiv.1702.06818,
  title  = {Stochastic Approximation for Canonical Correlation Analysis},
  author = {Raman Arora and Teodor V. Marinov and Poorya Mianjy and Nathan Srebro},
  journal= {arXiv preprint arXiv:1702.06818},
  year   = {2018}
}
R2 v1 2026-06-22T18:25:20.288Z