added torch3 legacy

index.html

@@ -315,6 +315,13 @@ <h2>
 and MacOS X. Core packages also work on Windows.
 </p>
 
+<h2>
+<a name="legacy" class="anchor" href="#legacy"><span class="octicon octicon-link"></span></a>Legacy
+<hr></h2>
+
+<p>
+The very old Torch3 version is still available <a href="torch3/index.html">here</a>, should you need it.
+
 </section>
     </div>
     <!-- FOOTER  -->

torch3/ChangeLog

@@ -0,0 +1,149 @@
+===============================================================================
+	Torch 3.1
+===============================================================================
+
+core
+====
+        * Bagging.cc
+                - Initialize "is_selected_examples" to NULL.
+
+        * ClassFormatDataSet.cc
+                - Convert all the frames of targets, and not only the
+                first one.
+
+        * ClassMeasurer.cc
+                - Scan all the frames of inputs to compute classification
+                error, and not only the first one.
+                - The confusion matrix computation option at the end of
+		the training as been removed. The confusion matrix is still
+		available at each iteration.
+
+        * CmdOption.cc
+                - When calling loadXFile(), a string not always allocated
+                was freed.
+
+        * DiskDataSet.cc
+                - The number of examples was computed using io_inputs, even
+                if it was NULL.
+
+        * KFold.cc
+                - The size of the folds is more balanced for small folds.
+
+	* MemoryDataSet.cc
+                - The number of examples was computed using io_inputs, even
+                if it was NULL.
+
+	* MemoryXFile.cc
+	        - The EOF flag was not updated when reaching the end of file
+		while reading with the scanf method.
+
+        * MultiClassFormat.cc
+                - Serious bug when auto-detecting classes corrected.
+
+        * Random.cc
+                - Using now memmove instead of memcopy (possibility of
+		overlapping source and destination).
+
+matrix
+======
+        * Mat.cc and Mat.h
+                - Initializing a matrix from a "real *" now possible.
+
+gradients
+=========
+
+        * SumMachine.cc
+                - The gradient is now well back-propagated.
+
+kernels
+=======
+        * QCTrainer.cc
+                - Bug when updating alpha corrected. Note: very rare
+                case in practice. Thanks to Stephen Schiller for the
+                report of this tricky bug.
+
+distributions
+=============
+       * DiagonalGMM.cc
+                - log_probabilities are now correctely computed for the viterbi case
+                - frameExpectation becomes frameDecision
+                - keep the best gaussian for the current frame
+
+       * Distribution.cc
+                - resize the output of the machine correctely
+                - added decision and frameDecision method
+
+       * HMM.cc
+                - added states shared parameters
+
+       * MAPDiagonalGMM.cc
+                - adapt the parameters only if the gaussian have "seen" at
+                  least one frame
+
+       * Multinomial.cc
+                - added equal initialization option
+                - added check that the log weights are number
+                - frameExpectation becomes frameDecision
+
+       * NLLCriterion.cc
+                - correct the beta and outputs resize
+
+       * ParzenDistribution.cc
+                - correct the resize of the log_probablities
+                - added frameExpectation method
+
+       * TableLookupDistribution.cc
+                - correction of some bugs to calculate the log probablity
+
+       * ViterbiTrainer.cc ,ViterbiTrainer.h
+                - this class is suppressed, it was already included in EMTrainer
+
+
+datasets
+========
+
+      * IOHTK.cc
+               - correction of some major bugs when reading the HTK files on
+                 disk on unsequential mode
+               - correction of major bug on double mode
+
+      * IOHTKTarget.cc
+               - space followed by tabulation allowed instead of space only
+               - added some check
+
+      * IOHTKTarget.h
+               - make saveSequence static
+
+      * Vocabulary.cc
+               - bound check corrected for the number of words
+
+speech
+======
+
+      * EditDistance.{cc,h} and EditDistanceMeasurer.{cc,h}
+               - Add a constructor option to print a confusion matrix as well
+
+      * WordSeg.{cc,h} and FrameSeg.{cc,h}
+               - Two new class to keep the word and frame segmentations found
+                 by viterbi decoding. These are kept separate so as to be
+                 compatible with any new subclass of SpeechHMM.
+
+      * WordSegMeasurer.{cc,h} and FrameSegMeasurer.{cc,h}
+               - Measurers that outputs information related to the number
+                 of errors in terms of word error rate (WordSegMeasurer) or
+                 frame error rate (FrameSegMeasurer), using the EditDistance
+                 class to compute the error.
+
+      * SimpleDecoderSpeechHMM.{cc,h}
+               - Modifications to take into account the new WordSeg and FrameSeg
+                 classes
+               - added a structure (previous_states[] and n_previous_states)
+                 to speedup the decoding process by only looping on existing
+                 transitions and not all possible transitions
+               - Added an option to be able to decode using "forced alignment",
+                 hence instead of decoding on the whole grammar, decoding on
+                 a given (true) sentence
+
+      * SpeechHMM.cc
+               - bug corrections related to initialization
+

torch3/credits.html

@@ -0,0 +1,65 @@
+<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
+<html>
+<head>
+  <meta http-equiv="Content-Type" content="text/html; charset=iso-8859-1">
+  <meta http-equiv="Content-Style-Type" content="text/css">
+  <title>Torch3' Credits</title>
+  <link rel="stylesheet" type="text/css" href="torch.css">
+  <link rel="icon" href="images/icon_logo.gif" type="image/ico">
+</head>
+
+<body>
+<table width="100%" cellpadding="0" cellspacing="0" border="0" cols="7">
+  <tr>
+    <td valign="middle" background="images/title_line2.jpg"><img src="images/logo.gif" width="78" height="71" alt="Torch3"></td>
+    <td align="center" valign="middle" background="images/title_line2.jpg" height="71"><br>&nbsp;&nbsp;<a href="introduction.html">Introduction</a>&nbsp;&nbsp;</td>
+    <td align="center" valign="middle" background="images/title_line2.jpg" height="71"><br>&nbsp;&nbsp;<a href="documentation.html">Documentation</a>&nbsp;&nbsp;</td>
+    <td align="center" valign="middle" background="images/title_line2.jpg" height="71"><br>&nbsp;&nbsp;<a href="downloads.html">Downloads</a>&nbsp;&nbsp;</td>
+    <td align="center" valign="middle" background="images/title_line2.jpg" height="71"><br>&nbsp;&nbsp;<a href="forum.html">Forum</a>&nbsp;&nbsp;</td>
+    <td align="center" valign="middle" background="images/title_line2.jpg" height="71"><br>&nbsp;&nbsp;<a href="credits.html">Credits</a>&nbsp;&nbsp;</td>
+  </tr>
+</table>
+<div align="right">       
+<small><i>
+<a href="http://www.idiap.ch/%7Ecollober">Ronan Collobert</a> 
+(<a href="mailto:collober [at] idiap.ch">collober [at] idiap.ch</a>)<br>
+</i>
+</small>
+</div>
+
+<h2>Authors</h2>
+  <ul>
+    <li> <a href="http://www.idiap.ch/~collober">Ronan Collobert</a></li>
+    <li> <a href="http://bengio.abracadoudou.com">Samy Bengio</a></li>
+    <li> <a href="http://www.idiap.ch/~marietho">Johnny Mari&eacute;thoz</a></li>
+  </ul>
+<hr width="100%">
+
+<h2>Contributors</h2>
+  <ul>
+    <li> <a href="http://www.idiap.ch/~gilardi">Nicolas Gilardi</a> for all the nice pictures he provided.</li>
+    <li> <a href="mailto:monay [at] idiap.ch">Florent Monay</a> for the <i><b>Torch</i></b> logo.</li>
+    <li> <a href="http://www.idiap.ch/~moore/">Darren Moore</a> for his speech decoder.</li>
+  </ul>
+<hr width="100%">
+
+<h2>Cool Stuff</h2>
+  <ul>
+    <li> <a href="http://www.phpBB.com/">phpBB</a> team for its very nice GPL'ed forum toolkit. <a href="http://www.phpBB.com" target="_blanck"><img align="middle" src="images/phpBB.gif" width="88" height="31" alt="Powered by phpBB2" title="Powered by phpBB2" border="0"></a></li>
+    <li> <a href="mailto:jseward [at] acm.org">Julian Seward</a> for his memory debugger tool, <a href="http://developer.kde.org/~sewardj/">Valgrind</a>.
+    Without it, <i><b>Torch3</b></i> wouldn't have been developped so fast!</li>
+  </ul>
+<hr width="100%">
+
+<h2>Special</h2>
+  <ul>
+   <li> <a href="http://leon.bottou.com">L&eacute;on Bottou</a> and <a href="http://yann.lecun.com">Yann Le Cun</a>
+        for some ideas in gradient machines which are coming from their <i>SN</i> library (now available as the <a href="http://lush.sourceforge.net/">Lush</a> library) and from L&eacute;on's Phd thesis. </li>
+   <li> <a href="mailto:daniel.collobert [at] francetelecom.com">Daniel Collobert</a> (<a href="http://www.rd.francetelecom.com">France Telecom R&amp;D</a>), for his computer on which <b><i>Torch3</i></b> is born.</li>
+   <li> <a href="mailto:formaz [at] idiap.ch">Frank Formaz</a> for the URL of the website.</li>
+   <br><li> <a href="http://www.hp.com">HP</a> for the computers they provided to us. <a href="http://www.hp.com"><img align="middle" src="images/hp.jpg" width="37" height="31" alt="HP" border="0"></a></li>
+  </ul>
+<hr width="100%">
+
+</body>
+</html>

torch3/documentation.html

@@ -0,0 +1,76 @@
+<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
+<html>
+<head>
+  <meta http-equiv="Content-Type" content="text/html; charset=iso-8859-1">
+  <meta http-equiv="Content-Style-Type" content="text/css">
+  <title>Torch3' Documentation</title>
+  <link rel="stylesheet" type="text/css" href="torch.css">
+  <link rel="icon" href="images/icon_logo.gif" type="image/ico">
+</head>
+
+<body>
+<table width="100%" cellpadding="0" cellspacing="0" border="0" cols="7">
+  <tr>
+    <td valign="middle" background="images/title_line2.jpg"><img src="images/logo.gif" width="78" height="71" alt="Torch3"></td>
+    <td align="center" valign="middle" background="images/title_line2.jpg" height="71"><br>&nbsp;&nbsp;<a href="introduction.html">Introduction</a>&nbsp;&nbsp;</td>
+    <td align="center" valign="middle" background="images/title_line2.jpg" height="71"><br>&nbsp;&nbsp;<a href="documentation.html">Documentation</a>&nbsp;&nbsp;</td>
+    <td align="center" valign="middle" background="images/title_line2.jpg" height="71"><br>&nbsp;&nbsp;<a href="downloads.html">Downloads</a>&nbsp;&nbsp;</td>
+    <td align="center" valign="middle" background="images/title_line2.jpg" height="71"><br>&nbsp;&nbsp;<a href="forum.html">Forum</a>&nbsp;&nbsp;</td>
+    <td align="center" valign="middle" background="images/title_line2.jpg" height="71"><br>&nbsp;&nbsp;<a href="credits.html">Credits</a>&nbsp;&nbsp;</td>
+  </tr>
+</table>
+<div align="right">       
+<small><i>
+<a href="http://www.idiap.ch/%7Ecollober">Ronan Collobert</a> 
+(<a href="mailto:collober [at] idiap.ch">collober [at] idiap.ch</a>)<br>
+</i>
+</small>
+</div>
+
+<h2>Disclaimer</h2>
+<blockquote>
+<i><b>Torch3</b></i> is <i>not</i> compatible with <i><b>Torch2</b></i>. The structure is similar, but a lot of things have been
+<i>simplified</i>. There is no script to convert your <i><b>Torch2</b></i> code, but it should be easy to do it. You should use
+<i><b>Torch3</b></i> because it's <i>easier</i> to code with it, and because we provide <i>tutorials</i> for a quick start.
+</blockquote>
+<hr width="100%">
+
+<h2>Installation</h2>
+<blockquote>
+<i><b>Torch</b></i> is pretty easy to install under <i>Unix</i> and <i>Linux</i> systems, including <i>Mac OS X</i>. Before asking questions,
+<ul>
+<li>have a look <a href="matos/install_xmake.pdf">here</a> if you want to use <i>xmake</i> (<b>recommended</b>).
+<li>have a look <a href="matos/install.pdf">here</a> if you want to use <i>GNU Make</i>.
+</ul>
+Note that <i>xmake</i> is a fast (and easy) <i>python</i> script designed for compiling <i><b>Torch</b></i>. Makefiles are not needed anymore.<br>
+<br>
+ For people using <i>MS Windows</i>, two solutions are possible:
+<ul>
+  <li> Install <i>Linux</i> using the free <a href="http://www.debian.org">Debian</a> distribution which provides more than 8000 useful packages, and
+then go <a href="matos/install_xmake.pdf">here</a> (<b>recommended</b>). </li>
+  <li> Otherwise, if you really want to keep <i>MS Windows</i>, use instructions provided <a href="matos/w_install.pdf">here</a>.
+</ul>
+</blockquote>
+<hr width="100%">
+
+<h2>Tutorials</h2>
+<blockquote>
+Several tutorials are provided for the beginner in <i><b>Torch</b></i>.
+<ul>
+  <li> <a href="matos/tutorial.pdf">Tutorial on <i><b>Torch</b></i></a> written by myself.</li>
+  <li> <a href="matos/distributions.pdf">Tutorial on distributions</a> written by Samy Bengio.</li>
+  <li> <a href="matos/convolutions.pdf">Tutorial on convolutions</a> written by myself.</li>
+  <li> <a href="matos/decoder.pdf">Manual of the speech decoder</a> written by <a href="mailto:moore [at] idiap.ch"> Darren Moore</a>.</li>
+</ul>
+</blockquote>
+<hr width="100%">
+
+<h2>One-line documentation</h2>
+<blockquote>
+If you are looking for a class, check the <a href="manual/HIER.html">class hierarchy</a>.<br>
+If you want to see the documentation for all classes, go <a href="manual/index.html">there</a>.
+<ul>
+</blockquote>
+<hr width="100%">
+</body>
+</html>