File: introduction.html

package info (click to toggle)
torch 2-1
  • links: PTS
  • area: main
  • in suites: woody
  • size: 5,488 kB
  • ctags: 3,217
  • sloc: cpp: 14,272; makefile: 201
file content (112 lines) | stat: -rw-r--r-- 3,971 bytes parent folder | download
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
<!doctype html public "-//w3c//dtd html 4.0 transitional//en">
<html>
<head>
   <meta http-equiv="Content-Type" content="text/html; charset=iso-8859-1">
   <meta name="GENERATOR" content="Mozilla/4.77 [en] (X11; U; Linux 2.4.10 i686) [Netscape]">
   <meta name="Author" content="Ronan Collobert">
   <meta name="Description" content="Torch Library">
   <meta name="Keywords" content="Machine Learning">
   <title>Torch Introduction</title>
</head>
<body text="#000000" bgcolor="#FFFFFF" link="#0000EE" vlink="#551A8B" alink="#FF0000">

<center><img SRC="images/title.jpeg" ALT="[Torch]" NOSAVE height=137 width=220><img SRC="images/release.jpeg" NOSAVE height=18 width=77>
<p><img SRC="images/torche.jpeg" ALT="[XXXXX]" NOSAVE height=32 width=32></center>

<div align=right><i><font size=-1><a href="http://www.idiap.ch/~collober">Ronan
Collobert</a>,
<a href="mailto:collober@iro.umontreal.ca">collober@iro.umontreal.ca</a></font></i></div>

<center><table COLS=5 WIDTH="100%" BGCOLOR="#FFCC00" NOSAVE >
<tr ALIGN=CENTER VALIGN=CENTER NOSAVE>
<td NOWRAP NOSAVE><a href="about.html"><img SRC="images/about_b.jpeg" ALT="[About]" NOSAVE BORDER=0 height=32 width=60></a></td>

<td NOWRAP NOSAVE><a href="introduction.html"><img SRC="images/introduction_bx.jpeg" ALT="[Introduction]" NOSAVE BORDER=0 height=32 width=95></a></td>

<td NOWRAP NOSAVE><a href="docs.html"><img SRC="images/docs_b.jpeg" ALT="[Docs]" NOSAVE BORDER=0 height=32 width=52></a></td>

<td NOWRAP NOSAVE><a href="install.html"><img SRC="images/install_b.jpeg" ALT="[Install]" NOSAVE BORDER=0 height=32 width=65></a></td>

<td NOWRAP NOSAVE><a href="downloads.html"><img SRC="images/downloads_b.jpeg" ALT="[Downloads]" NOSAVE BORDER=0 height=32 width=89></a></td>
</tr>
</table></center>

<p><img SRC="images/new.jpg" ALT="[new!]" NOSAVE height=11 width=28> You
should have a look on <a href="matos/presentation.ps.gz">this presentation</a>
written by <i><a href="http://www.idiap.ch/~bengio">Samy Bengio</a></i>
which introduces main concepts in <b><i>Torch</i></b>.
<p>Here is an abstract:
<br>There are several concepts to know in <b><i>Torch</i></b>, if you plan
to use it:
<ul>
<li>
DataSet</li>

<li>
Machine</li>

<li>
Trainer</li>

<li>
Measurer</li>
</ul>

<p><br>A <i>DataSet</i> is a black-box which could provide all kind of
data. For example you have:
<ul>
<li>
StdDataSet, which manages standard datasets: examples with the same input
dimension and the same output dimension.</li>

<li>
SeqDataSet which provides a easy interface to sequence datasets used in
the Distribution package...</li>
</ul>
A <i>Machine </i>is a learning-machine. It could be:
<ul>
<li>
MLP, a multi-layered perceptron.</li>

<li>
HMM, a hidden markov model</li>

<li>
SVM, a support vector machine</li>
</ul>
A <i>Trainer</i> is a class which takes a <i>Machine</i> and a <i>DataSet</i>,
and which is able to train or test the machine. It could be:
<ul>
<li>
GMTrainer, a gradient machine trainer which can train all kind of gradient
machines. (For example MLP)</li>

<li>
EMTrainer, which trains several machines using the EM algorithm. (For example
HMM)</li>

<li>
QCTrainer, which trains all quadratic, constrained machine. (For example
SVM)</li>
</ul>
A <i>Measurer </i>is a special class which is called by a <i>Trainer</i>,
during the train phase or the test phase. It's always associated to a <i>DataSet</i>.
It could measure everything on your machine, on the <i>DataSet</i> that
you provide. Here are some examples:
<ul>
<li>
MseMeasurer, which measures the mean squared error of you machine, on the
<i>DataSet</i>.</li>

<li>
ClassMeasurer, which measures the classification error when you have a
classification problem.</li>
</ul>

<p><br>Ok, almost all classes in <b><i>Torch </i></b>are one of these classes...
Easy, no ? If you want to have an idea of all classes which are available
to the public, just have a look on the <a href="manual/index.html">reference
manual</a>...
<br>&nbsp;
</body>
</html>