annotate DL/Majorization Minimization DL/README @ 220:0d30f9074dd9

Merge
author luisf <luis.figueira@eecs.qmul.ac.uk>
date Wed, 11 Apr 2012 15:56:39 +0100
parents b14209313ba4
children
rev   line source
ivan@155 1 Majorization Minimization Dictionary Learning (MMDL) Package (Beta version)
ivan@155 2
ivan@155 3 This package includes the codes needed to generate some of the plots in the paper, "Dictionary Learning for Sparse Approximations with the Majorization Method" by M. Yaghoobi, T. Blumensath and M. Davies, IEEE Transaction on Signal Processing, Vol. 57, No. 6, pp 2178-2191, 2009.
ivan@155 4
ivan@155 5 The package also includes the authors' implementations of some other well-known dictionary learning methods. These have been used for comparison reason.
ivan@155 6
ivan@155 7 The package includes two parts:
ivan@155 8
ivan@155 9 1- Building blocks of the MMDL:
ivan@155 10
ivan@155 11 a) mm1 : Iterative Softthresholding for sparse approximation
ivan@155 12 b) rdlcn : MMDL with column norm constraint
ivan@155 13 c) rdlfn : MMDL with Frobenius norm constraint
ivan@155 14 d) softthreshold : Softthresholding operator
ivan@155 15 e) Demo : A demo to demonstrate how to use MMDL and recover a generic dictionary
ivan@155 16
ivan@155 17 2- MMDL in action:
ivan@155 18 a) ExactDicoRecovery : Exact generic atom recovery experiment. It includes a comparison with other methods
ivan@155 19 b) Dictionary Learning for Audio : Dictionary learning with MMDL method applied to some classic music recorded from BBC radio 3
ivan@155 20
ivan@155 21 Installation:
ivan@155 22
ivan@155 23 You just need to decompress the package into your local disk and add the path of main directory to the Matlab path.
ivan@155 24
ivan@155 25 Citation:
ivan@155 26
ivan@155 27 This package is free to use. However, any publication based on using this package, in any kind, should cite the mentioned paper.
ivan@155 28
ivan@155 29 Copyright:
ivan@155 30
ivan@155 31 The use of this package does not need any farther permission, if it is restricted to the academic and industrial researches. Any military affiliated use of this package is strictly prohibited.