📄 readme.html
字号:
<!DOCTYPE doctype PUBLIC "-//w3c//dtd html 4.0 transitional//en">
<html>
<head>
<meta http-equiv="Content-Type"
content="text/html; charset=iso-8859-1">
<meta name="Author" content="Luis B. Almeida">
<meta name="GENERATOR"
content="Microsoft FrontPage 4.0">
<title>MI-Based Linear and Nonlinear ICA Toolbox</title>
</head>
<body>
<center>
<p><b><font size="+1">MISEP Linear and Nonlinear ICA Toolbox<br>
</font></b></p>
<p><b><font size="+1"><small>Version 1.2</small><br>
</font></b></p>
</center>
<p><span style="font-weight: bold;">0. What's new in this version<br>
</span></p>
<p>Version 1.2 introduced two kinds of naming changes, to make the toolbox
compatible with Matlab 7:</p>
<ul>
<li>All filenames were changed to lowercase. This eliminated a number of
warnings that Matlab 7 generated (version 7 of Matlab is sensitive to
upper/lowercase in filenames, contrary to what happened with previous
versions).<br>
<br>
</li>
<li>The script named 'report' was changed to 'reportresults', to eliminate a
conflict with a function named 'report' that appeared in Matlab version 7
and didn't exist in previous versions of Matlab.<br>
<br>
</li>
</ul>
<h3> 1. Introduction</h3>
This page briefly describes a MATLAB toolbox for Independent Component
Analysis (ICA) and Blind Source Separation (BSS), based on the MISEP
method.
<p>The toolbox can perform both linear and nonlinear ICA, and is named 'MIToolbox'.. <br>
</p>
<h3> 2. Toolbox download</h3>
The toolbox can be downloaded <a href="http://neural.inesc-id.pt/~lba/ICA/MIToolbox.zip">here</a> (zip
file, around 50 kBytes). <br>
<h3> 3. Installation</h3>
To install the toolbox simply unzip all the files contained in the zip
archive into a directory of your choice. <br>
<h3> 4. Documentation</h3>
<h4> 4.1 ICA method</h4>
Papers that describe the ICA method can be downloaded <a href="http://neural.inesc-id.pt/~lba//research.html#misep">here</a>. <br>
<h4> 4.2 Detailed documentation</h4>
A manual is included with the toolbox. <br>
<h4> 4.3 Brief usage instructions</h4>
<ol>
<li><big> <font size="-1"><big>In MATLAB, '<tt>cd</tt>' to the
directory containing the toolbox files.</big></font></big></li>
<br>
<li><big> <font size="-1"><big>Give the command '<tt>netpar</tt>'
(this initializes the toolbox parameters).</big></font></big></li>
<br>
<li><big> <font size="-1"><big>Create an array named <tt>'trpattern'</tt>
with the data to be analyzed.</big></font></big></li>
<big><br>
<font size="-1"><big>The array should have one mixture component per
row and one mixture pattern (mixture vector) per column.</big></font> <br>
<font size="-1"><big>Give the commands:</big></font></big>
<ol>
<big><tt><font size="-1"><big>ninputs = <number of mixture
components></big></font></tt> <br>
<tt><font size="-1"><big>ntrain = <number of training
patterns></big></font></tt> <br>
</big>
</ol>
<big><font size="-1"><big>Note: For a quick test of the toolbox just
give the command '<tt>generate'</tt>, which will generate a nonlinear
mixture of two supergaussian random sources, with <tt>ntrain</tt>
training patterns. The variable <tt>'ntrain'</tt> is initially set to
1000, by the <tt>'netpar'</tt> command.</big></font> </big><br>
<li><big> <font size="-1"><big>Give the command '<tt>netinit</tt>'
(this initializes the analysis network's weights and auxiliary data).</big></font></big></li>
<br>
<li><big> <font size="-1"><big>Give the command '<tt>train</tt>'.</big></font></big></li>
</ol>
<big><font size="-1"><big><br>
Training should start. The value of the cost function, the cost
function improvement and epoch number should start being reported, one
epoch per line.</big></font></big>
<p><big><font size="-1"><big>A figure should be displayed, and should
be updated once every five epochs. Resize this figure so that the left
and center plots are square. The figure shows:</big></font> </big></p>
<ul>
<big> </big><li><big> <font size="-1"><big>Left: Scatter plot of the
extracted components (1<sup>st</sup> and 2<sup>nd</sup> components only).</big></font></big></li>
<big> </big><li><big> <font size="-1"><big>Middle: Scatter plot of
the auxiliary (<i>z</i>) outputs (1<sup>st</sup> and 2<sup>nd</sup>
components only).</big></font></big></li>
<big> </big><li><big> <font size="-1"><big>Right: Estimated
cumulative probability functions of the extracted components(1<sup>st</sup>
and 2<sup>nd</sup> components only, scaled to [-1,1] instead of [0,1]).</big></font></big></li>
<big> </big>
</ul>
<big><font size="-1"><big>If you want to stop the training before
reaching the set number of training epochs, press Ctrl-C (you may also
set a desired number of training epochs by setting the variable '<tt>nepochs</tt>'
before giving the <tt>train</tt> command).</big></font></big>
<p><big><font size="-1"><big>To process test data with the resulting
ICA system:</big></font> </big></p>
<ol>
<big> </big><li><big> <font size="-1"><big>Create an array named '<tt>mixeddata'</tt>
with the mixture to be processed. The format is the same as for '<tt>trpattern</tt>'
above, except that the number of columns doesn't need to be equal to <tt>'ntrain'</tt>.</big></font></big></li>
<big> </big><li><big> <font size="-1"><big>Give the command '<tt>processdata</tt>'.</big></font></big></li>
<big> </big><li><big> <font size="-1"><big>The extracted components
will be available in an array called '<tt>separdata</tt>', with the same
format as '<tt>mixeddata'</tt>.</big></font></big></li>
<big> </big>
</ol>
<h3> 5. Contact</h3>
Toolbox author:
<p>Luis B. Almeida <br>
INESC-ID <br>
R. Alves Redol, 9 <br>
1000-029 Lisboa <br>
Portugal </p>
<p>E-mail (delete the first two c's, which are there to prevent
spamming: <a href="mailto:ccluis.almeida@inesc-id.pt">ccluis.almeida@inesc-id.pt</a> </p>
<p>Home page: <a href="http://neural.inesc.pt/%7Elba/">http://neural.inesc.pt/~lba/</a> <br>
</p>
<h3> 6. Acknowledgement</h3>
The development of this toolbox has been partially supported by the
European IST project BLISS and by Praxis project P/EEI/14091/1998.
<p> </p>
<br>
<br>
</body>
</html>
⌨️ 快捷键说明
复制代码
Ctrl + C
搜索代码
Ctrl + F
全屏模式
F11
切换主题
Ctrl + Shift + D
显示快捷键
?
增大字号
Ctrl + =
减小字号
Ctrl + -