focuss
所属分类:matlab编程
开发工具:matlab
文件大小:39KB
下载次数:15
上传日期:2015-12-29 10:18:52
上 传 者:
Lx941314
说明: 利用FOCUSS算法对数据进行重构表示,得到需要表示的成分,减少噪声干扰。
(FOCUSS algorithms use the data to reconstruct said component required to obtain representation, reducing noise.)
文件列表:
focuss2\calcsnr.m (1104, 2003-08-29)
focuss2\compareframe.m (1670, 2001-04-12)
focuss2\comparesolution.m (1361, 2000-12-04)
focuss2\convertimages.m (3020, 2004-01-08)
focuss2\createdata.m (2558, 2002-05-10)
focuss2\createimagedata.m (1144, 2004-02-19)
focuss2\entropy.m (661, 2002-06-20)
focuss2\focuss.m (4125, 2014-10-05)
focuss2\learnrd.m (5676, 2004-11-29)
focuss2\license.txt (2124, 2005-02-15)
focuss2\numerosity.m (644, 2003-08-28)
focuss2\pickhighest.m (442, 2000-10-06)
focuss2\plotAhistory.m (841, 2003-10-24)
focuss2\plothistogram.m (827, 2001-03-22)
focuss2\plotresults.m (10236, 2004-12-02)
focuss2\plotresultsimage.m (9664, 2003-09-04)
focuss2\reorderica.m (776, 2002-01-21)
focuss2\runtrials_focusscndl.m (6990, 2004-11-29)
focuss2\runtrials_image.m (6643, 2004-02-19)
focuss2\sampleimage.m (2461, 2003-09-04)
focuss2\sortcolumns.m (1070, 2002-02-05)
focuss2\spheredata.m (728, 2001-01-20)
focuss2\trainrd_extica.m (1933, 2002-01-21)
focuss2\trainrd_fastica.m (1999, 2002-01-22)
focuss2\trainrd_focusscndl.m (13937, 2005-02-15)
focuss2\trainrd_focusscndl2.m (14243, 2005-02-15)
focuss2\trainrd_focussdl.m (12403, 2005-02-15)
focuss2\whitendata.m (1671, 2002-01-17)
focuss2 (0, 2014-12-04)
FOCUSS-CNDL Dictionary Learning Algorithms for Sparse Representations
Matlab Code Documentation
Contact:
Joseph Murray
jfmurray@ucsd.edu
http://dsp.ucsd.edu/~jfmurray
Kenneth Kreutz-Delgado
kreutz@ece.ucsd.edu
http://dsp.ucsd.edu/~kreutz
Copyright 2005 Joseph F. Murray
---- Overview ----
The Matlab code in this directory implements the FOCUSS-CNDL learning
algorithm from the paper,
@ARTICLE{Kreutz:2003,
author = {Kenneth Kreutz-Delgado and Joseph F. Murray and Bhaskar D. Rao and Kjersti Engan and Te-Won Lee and Terrence J. Sejnowski},
title = {Dictionary Learning Algorithms for Sparse Representation},
journal = {Neural Computation},
year = {2003},
volume = {15},
number = {2},
pages = {349-396},
month = {February}
}
This code is still in a rough, experimental form. Please contact us (see
above) if you have any questions or difficulties.
The software is distributed according to the license described in the
license.txt file included with this distribution.
---- Learning Image Dictionaries ----
Start by creating a data set of patches drawn from the images, using
createimagedata.m. This loads all the images in a specified directory
draws small patches at random, and saves them to a .mat file.
The dictionaries are learned based on this .mat data by runtrials_image.m,
which calls the actual learning algorithm, found in trainrd_focusscndl.m
or trainrd_focusscndl2.m (both should give similar results).
Once a dictionary has been learned, results can be plotted with
plotresultsimage.m, which also calculates the entropy of the learned
coefficients using a method of similar to Lewicki:1999.
A stand-alone version of the FOCUSS algorithm, focuss.m, can be used to
find the coding of a new data set once the dictionary is learned.
---- Synthetic Dictionaries ----
The 20x30 synthetic dictionary experiment is run using
runtrials_focusscndl.m, which generates random dictionaries.
Results are plotted with plotresults.m, which compares the learned
dictionary to the true generating dictionary.
近期下载者:
相关文件:
收藏者: