ICA Using Overcomplete Representations

  • Te-Won Lee
Chapter

Abstract

In this chapter empirical results are shown for the blind source separation of more sources than mixtures using a framework proposed for learning overcomplete representations recently developed by Lewicki and Sejnowski (1998b).

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

Notes

  1. 1.
    In most ICA formulations, the matrix A is restricted M N,which is not imposed here.Google Scholar
  2. 2.
    The original, mixed and inferred signals are available in http://www.cnl.Salk.edu/~tewon/.

Copyright information

© Springer Science+Business Media Dordrecht 1998

Authors and Affiliations

  • Te-Won Lee
    • 1
  1. 1.Computational Neurobiology LaboratoryThe Salk InstituteLa JollaUSA

Personalised recommendations