Kernel-based low-rank feature extraction on a budget for big data streams

Fatemeh Sheikholeslami, Dimitris Berberidis, Georgios B. Giannakis

Research output: Chapter in Book/Report/Conference proceedingConference contribution

7 Scopus citations

Abstract

With nowadays big data torrent, identifying low-dimensional latent structures and extracting features from massive datasets are tasks of paramount importance. To this end, as real data generally lie on (or close to) nonlinear manifolds, kernel-based approaches are well motivated. Being nonparametric, unfortunately kernel-based feature extraction incurs complexity that grows prohibitively with the number of data. In response to this formidable challenge, the present work puts forward a low-rank, kernel-based feature extraction method, where the number of kernel functions is confined to an affordable budget. The resultant algorithm is particularly tailored for online operation, where data streams need not even be stored in memory. Tests on synthetic and real datasets demonstrate and benchmark the efficiency of the proposed method on linear classification applied to the extracted features.

Original languageEnglish (US)
Title of host publication2015 IEEE Global Conference on Signal and Information Processing, GlobalSIP 2015
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages928-932
Number of pages5
ISBN (Electronic)9781479975914
DOIs
StatePublished - Feb 23 2016
EventIEEE Global Conference on Signal and Information Processing, GlobalSIP 2015 - Orlando, United States
Duration: Dec 13 2015Dec 16 2015

Publication series

Name2015 IEEE Global Conference on Signal and Information Processing, GlobalSIP 2015

Other

OtherIEEE Global Conference on Signal and Information Processing, GlobalSIP 2015
Country/TerritoryUnited States
CityOrlando
Period12/13/1512/16/15

Bibliographical note

Funding Information:
Work in this paper was supported by NSF grants, 1343248, 1500713, and NIH grant no. 1R01GM104975-01.

Keywords

  • budgeted learning
  • classification
  • kernel methods
  • online feature extraction
  • subspace tracking

Fingerprint

Dive into the research topics of 'Kernel-based low-rank feature extraction on a budget for big data streams'. Together they form a unique fingerprint.

Cite this