Change search
ReferencesLink to record
Permanent link

Direct link
SAGI: Sparsification Algorithm using Greedy Iteration
University College Dublin.
University College Dublin, Ireland.ORCID iD: 0000-0002-3912-1470
Neural Engineering Group, Trinity Centre for Bioengineering.
2005 (English)In: Signal Processing with Adaptive Sparse Structured Representations, Rennes, 2005, 16-8 p.Conference paper (Refereed)
Abstract [en]

We introduce a method, called SAGI (Sparsification Algo- rithm using Greedy Iteration), for making a representation of a signal more sparse in an over-complete dictionary in a greedy manner. The sparsification is achieved by iteratively increasing the magnitude of the largest signal coefficient and simultaneously reducing the other signal coefficients so as to maximize the sparsity of the representation while main- taining invariant the reconstruction of the signal from the coefficients. Any measure of sparsity can be used with the method. Two versions are presented. The first, the one-pass version, considers each coefficient once in order from largest to smallest. The second, the exhaustive version, only consid- ers the (n + 1)st largest coefficient if iteratively considering the n largest coefficients results in no increase in sparsity.

Place, publisher, year, edition, pages
Rennes, 2005. 16-8 p.
Keyword [en]
Sparse Signal Processing; Optimization; Blind Source Separation
National Category
Signal Processing
Research subject
Applied and Computational Mathematics
URN: urn:nbn:se:kth:diva-174171OAI: diva2:860981
Workshop on Signal Processing with Adaptive Sparse Structured Representations (SPARS'05)

QC 20151015

Available from: 2015-10-14 Created: 2015-10-01 Last updated: 2015-10-15Bibliographically approved

Open Access in DiVA

No full text

Other links

Search in DiVA

By author/editor
de Fréin, Ruairí
Signal Processing

Search outside of DiVA

GoogleGoogle Scholar
The number of downloads is the sum of all downloads of full texts. It may include eg previous versions that are now no longer available

Total: 20 hits
ReferencesLink to record
Permanent link

Direct link