1 libDAI - A free/open source C++ library for Discrete Approximate Inference methods

2 ==================================================================================

4 v 0.2.2 - September 30, 2008

6 ----------------------------------------------------------------------------------

8 This file is part of libDAI - http://www.libdai.org/

10 libDAI is free software; you can redistribute it and/or modify

11 it under the terms of the GNU General Public License as published by

12 the Free Software Foundation; either version 2 of the License, or

13 (at your option) any later version.

15 libDAI is distributed in the hope that it will be useful,

16 but WITHOUT ANY WARRANTY; without even the implied warranty of

17 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the

18 GNU General Public License for more details.

20 You should have received a copy of the GNU General Public License

21 along with libDAI in the file COPYING.

22 If not, see http://www.gnu.org/licenses/

24 ----------------------------------------------------------------------------------

26 Copyright (C) 2006-2009 Joris Mooij [joris dot mooij at libdai dot org]

27 Copyright (C) 2002-2007 Radboud University Nijmegen, The Netherlands

28 Copyright (C) 2002 Martijn Leisink [martijn@mbfys.kun.nl]

30 with contributions from:

32 Martijn Leisink

33 Giuseppe Passino

34 Frederik Eaton

35 Charlie Vaske

36 Bastian Wemmenhove

37 Christian Wojek

38 Claudio Lima

39 Jiuxiang Hu

40 Peter Gober

41 Patrick Pletscher

42 Sebastian Nowozin

44 ----------------------------------------------------------------------------------

46 If you write a scientific paper describing research that made substantive use

47 of this program, please (a) mention the fashion in which this software was

48 used, including the version number, with a citation to the literature, to allow

49 replication; (b) mention this software in the Acknowledgements section. The

50 appropriate citation is:

52 J. M. Mooij (2008) "libDAI 0.2.2: A free/open source C++ library for Discrete

53 Approximate Inference methods", http://www.libdai.org

55 Moreover, as a personal note, I would appreciate it if you would email me

56 (joris.mooij@libdai.org) with citations of papers referencing this work.

58 ----------------------------------------------------------------------------------

61 About libDAI

62 ------------

63 libDAI is a free/open source C++ library (licensed under GPL) that provides

64 implementations of various (approximate) inference methods for discrete

65 graphical models. libDAI supports arbitrary factor graphs with discrete

66 variables; this includes discrete Markov Random Fields and Bayesian Networks.

68 The library is targeted at researchers; to be able to use the library, a good

69 understanding of graphical models is needed.

72 Limitations

73 -----------

74 libDAI is not intended to be a complete package for approximate inference.

75 Instead, it should be considered as an "inference engine", providing various

76 inference methods. In particular, it contains no GUI, currently only supports

77 its own file format for input and output (although support for standard file

78 formats may be added later), and provides very limited visualization

79 functionalities.

82 Features

83 --------

84 Currently, libDAI supports the following (approximate) inference methods:

86 * Exact inference by brute force enumeration;

87 * Exact inference by junction-tree methods;

88 * Mean Field;

89 * Loopy Belief Propagation [KFL01];

90 * Tree Expectation Propagation [MiQ04];

91 * Generalized Belief Propagation [YFW05];

92 * Double-loop GBP [HAK03];

93 * Various variants of Loop Corrected Belief Propagation [MoK07, MoR05];

94 * Gibbs sampler;

95 * Conditioned BP [EaG09].

97 These inference methods can be used to calculate partition sums, marginals

98 over subsets of variables, and MAP states (the joint state of variables that

99 has maximum probability).

101 In addition, libDAI supports parameter learning of conditional probability

102 tables by Expectation Maximization.

105 Why C++?

106 --------

107 Because libDAI is implemented in C++, it is very fast compared with

108 implementations in MatLab (a factor 1000 faster is not uncommon). libDAI does

109 provide a (limited) MatLab interface for easy integration with MatLab.

112 Releases

113 --------

114 Releases can be obtained from www.libdai.org

115 License: GNU Public License v2 (or higher).

117 libDAI-0.2 December 1, 2006

118 libDAI-0.2.1 May 26, 2008

119 libDAI-0.2.2 September 30, 2008

122 Acknowledgments

123 ---------------

124 This work is part of the Interactive Collaborative Information Systems (ICIS)

125 project, supported by the Dutch Ministry of Economic Affairs, grant BSIK03024.

126 I would like to thank Martijn Leisink for providing the basis on which libDAI has been built.

129 Documentation

130 -------------

131 Some doxygen documentation is available. Install doxygen and use "make doc" to build the

132 documentation. If the documentation is not clear enough, feel free to send me an email

133 (or even better, to improve the documentation!).

135 A description of the factor graph (.fg) file format can be found in the file FILEFORMAT.

138 Compatibility

139 -------------

140 The code has been developed under Debian GNU/Linux with the GCC compiler suite.

141 libDAI compiles successfully with g++ versions 3.4, 4.1, 4.2 and 4.3.

143 libDAI has also been successfully compiled with MS Visual Studio 2008 under Windows

144 (but not all build targets are supported yet) and with Cygwin under Windows.

146 Finally, libDAI has been compiled successfully on MacOS X.

149 Quick start (linux/cygwin/Mac OS X)

150 -----------------------------------

151 You need:

152 - a recent version of gcc (at least version 3.4)

153 - GNU make

154 - doxygen

155 - graphviz

156 - recent boost C++ libraries (at least version 1.34, or 1.37 for cygwin;

157 version 1.37 shipped with Ubuntu 9.04 is known not to work)

159 On Debian/Ubuntu, you can easily install all these packages with a single command:

160 "apt-get install g++ make doxygen graphviz libboost-dev libboost-graph-dev libboost-program-options-dev"

161 (root permissions needed).

163 On Mac OS X (10.4 is known to work), these packages can be installed easily via MacPorts.

164 First, install MacPorts according to the instructions at http://www.macports.org/

165 Then, a simple "sudo port install gmake boost doxygen graphviz"

166 should be enough to install everything that is needed.

168 On Cygwin, the prebuilt Cygwin package boost-1.33.1-x is known not to work.

169 You can however obtain the latest boost version (you need at least 1.37.0)

170 from http://www.boost.org/ and compile/install it with:

172 ./configure

173 make

174 make install

177 To build the libDAI source, first copy a template Makefile.* to Makefile.conf

178 (for example, copy Makefile.LINUX to Makefile.conf if you use GNU/Linux).

179 Then, edit the Makefile.conf template to adapt it to your local setup.

180 Especially directories may differ from system to system. Finally, run

182 make

184 If the build was successful, you can test the example program:

186 ./example tests/alarm.fg

188 or the more elaborate test program:

190 tests/testdai --aliases tests/aliases.conf --filename tests/alarm.fg --methods JTREE_HUGIN BP_SEQMAX

193 Quick start (Windows)

194 ---------------------

195 You need:

196 - A recent version of MicroSoft Visual Studio (2008 works)

197 - recent boost C++ libraries (version 1.34 or higher)

198 - GNU make (can be obtained from http://gnuwin32.sourceforge.net)

199 For the regression test, you need:

200 - GNU diff, GNU sed (can be obtained from http://gnuwin32.sourceforge.net)

202 To build the source, copy Makefile.WINDOWS to Makefile.conf. Then, edit

203 Makefile.conf to adapt it to your local setup. Finally, run (from the command line)

205 make

207 If the build was successful, you can test the example program:

209 example tests\alarm.fg

211 or the more elaborate test program:

213 tests\testdai --aliases tests\aliases.conf --filename tests\alarm.fg --methods JTREE_HUGIN BP_SEQMAX

216 Quick start (MatLab)

217 --------------------

218 You need:

219 - MatLab

220 - The platform-dependent requirements described above

222 First, you need to build the libDAI source as described above for your

223 platform. By default, the MatLab interface is disabled, so before compiling the

224 source, you have to enable it in the Makefile.conf by setting

225 "WITH_MATLAB=true". Also, you have to configure the MatLab-specific parts of

226 Makefile.conf to match your system (in particular, the Makefile variables ME,

227 MATLABDIR and MEX). The MEX file extension depends on your platform; for a

228 64-bit linux x86_64 system this would be "ME=.mexa64", for a 32-bit linux x86

229 system this would be "ME=.mexglx". If you are unsure about your MEX file

230 extension: it needs to be the same as what the MatLab command "mexext" returns.

231 The required MEX files are built by issuing

233 make

235 from the command line. The MatLab interface is much less powerful than using

236 libDAI from C++. There are two reasons for this: (i) it is boring to write MEX

237 files; (ii) the large performance penalty paid when large data structures (like

238 factor graphs) have to be converted between their native C++ data structure to

239 something that MatLab understands.

241 A simple example of how to use the MatLab interface is the following (entered

242 at the MatLab prompt), which performs exact inference by the junction tree

243 algorithm and approximate inference by belief propagation on the ALARM network:

245 cd path_to_libdai/matlab

246 [psi] = dai_readfg ('../examples/alarm.fg');

247 [logZ,q,md,qv,qf] = dai (psi, 'JTREE', '[updates=HUGIN,verbose=0]')

248 [logZ,q,md,qv,qf] = dai (psi, 'BP', '[updates=SEQMAX,tol=1e-9,maxiter=10000,logdomain=0]')

250 where "path_to_libdai" has to be replaced with the directory in which libDAI

251 was installed. For other algorithms and their parameters, see

252 tests/aliases.conf.