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

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

4 v 0.2.2 - May 26, 2008

7 Copyright (C) 2006-2008 Joris Mooij [joris dot mooij at tuebingen dot mpg dot de]

8 Radboud University Nijmegen, The Netherlands /

9 Max Planck Institute for Biological Cybernetics, Germany

12 ----------------------------------------------------------------------------------

13 This file is part of libDAI.

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

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

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

18 (at your option) any later version.

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

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

22 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the

23 GNU General Public License for more details.

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

26 along with libDAI; if not, write to the Free Software

27 Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA

28 ----------------------------------------------------------------------------------

31 SCIENTISTS: please be aware that the fact that this program is released as Free

32 Software does not excuse you from scientific propriety, which obligates you to

33 give appropriate credit! If you write a scientific paper describing research

34 that made substantive use of this program, it is your moral obligation as a

35 scientist to (a) mention the fashion in which this software was used, including

36 the version number, with a citation to the literature, to allow replication;

37 (b) mention this software in the Acknowledgements section. The appropriate

38 citation is: J. M. Mooij (2008) libDAI: A free/open source C++ library for

39 Discrete Approximate Inference methods, http://mloss.org/software/view/77/.

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

41 citations of papers referencing this work so I can mention them to my funding

42 agent and tenure committee.

44 What is libDAI?

45 ---------------

46 libDAI is a free/open source C++ library (licensed under GPL, see the file

47 COPYING for more details) that provides implementations of various

48 (deterministic) approximate inference methods for discrete graphical models.

49 libDAI supports arbitrary factor graphs with discrete variables (this includes

50 discrete Markov Random Fields and Bayesian Networks).

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

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

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

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

56 formats may be added), and provides no visualization.

58 Because libDAI is implemented in C++, it is very fast compared with e.g. MatLab

59 implementations. libDAI does provide a MatLab interface for easy integration

60 with MatLab. Currently, libDAI supports the following deterministic approximate

61 inference methods:

63 * Mean Field

64 * (Loopy) Belief Propagation

65 * Tree Expectation Propagation

66 * Generalized Belief Propagation

67 * Double-loop GBP

68 * Loop Corrected Approximate Inference

70 Exact inference by JunctionTree is also provided.

72 Many of these algorithms are not yet available in similar open source software,

73 to the best of the author's knowledge (open source packages supporting both

74 directed and undirected graphical models are Murphy's BNT, Intel's PNL and gR).

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

77 understanding of graphical models is needed. However, the code will hopefully

78 find its way into real-world applications as well.

81 Rationale

82 ---------

83 In my opinion, the lack of open source reference implementations hampers

84 progress in research on approximate inference. Methods differ widely in terms

85 of quality and performance characteristics, which also depend in different ways

86 on various properties of the graphical models. Finding the best approximate

87 inference method for a particular application therefore often requires

88 empirical comparisons. However, implementing and debugging these methods takes

89 a lot of time which could otherwise be spent on research. I hope that this code

90 will aid researchers to be able to easily compare various (existing as well as

91 new) approximate inference methods, in this way accelerating research and

92 stimulating real-world applications of approximate inference.

95 Releases

96 --------

97 Releases can be obtained from http://www.mbfys.ru.nl/~jorism/libDAI.

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

100 libDAI-0.2 December 1, 2006

101 libDAI-0.2.1 May 26, 2008

104 Acknowledgments

105 ---------------

106 The development reported here is part of the Interactive Collaborative

107 Information Systems (ICIS) project, supported by the Dutch Ministry of Economic

108 Affairs, grant BSIK03024. I would like to thank Martijn Leisink for providing

109 the basis on which libDAI has been built.

112 Known issues

113 ------------

114 Due to a bug in GCC 3.3.x and earlier (http://gcc.gnu.org/bugzilla/show_bug.cgi?id=20358)

115 it doesn't compile with these versions (it does compile with GCC version 3.4 and higher).

116 Workaround: replace the two NAN's in factor.h causing the error messages by -1.

119 Documentation

120 -------------

121 Almost nonexistant. But I'm working on it. In the meantime, I'll provide limited support

122 by email. The following gives an overview of different methods and their properties

123 (can be slightly obsolete):

125 BP

126 updates UpdateType SEQFIX,SEQRND,SEQMAX,PARALL

127 tol double

128 maxiter size_t

129 verbose size_t

130 MF

131 tol double

132 maxiter size_t

133 verbose size_t

134 HAK

135 clusters MIN,DELTA,LOOP

136 loopdepth

137 doubleloop bool

138 tol double

139 maxiter size_t

140 verbose size_t

141 JTREE

142 updates UpdateType HUGIN,SHSH

143 verbose size_t

144 MR

145 updates UpdateType FULL,LINEAR

146 inits InitType RESPPROP,CLAMPING,EXACT

147 verbose size_t

148 TREEEP

149 type TypeType ORG,ALT

150 tol double

151 maxiter size_t

152 verbose size_t

153 LC

154 cavity CavityType FULL,PAIR,PAIR2,UNIFORM

155 updates UpdateType SEQFIX,SEQRND(,NONE)

156 reinit bool

157 cavainame string

158 cavaiopts Properties

159 tol double

160 maxiter size_t

161 verbose size_t

165 Quick start

166 -----------

167 You need:

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

169 - the Boost C++ libraries (under Debian/Ubuntu you can install them using

170 "apt-get install libboost-dev libboost-program-options-dev")

171 - GNU make

173 To build the source, edit the Makefile and then run

175 make

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

179 ./example tests/alarm.fg

181 or the more elaborate test program:

183 tests/test --aliases tests/aliases.conf --filename tests/alarm.fg --methods JTREE_HUGIN BP_SEQMAX

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