[Peter Rockett] Improved Makefiles for image segmentation example.
[libdai.git] / README
1 libDAI - A free/open source C++ library for Discrete Approximate Inference
2
3 -------------------------------------------------------------------------------
4
5 Version: git HEAD
6 Date: May 12, 2010, or later
7 See also: http://www.libdai.org
8
9 -------------------------------------------------------------------------------
10
11 License
12
13 libDAI is free software; you can redistribute it and/or modify it under the
14 terms of the GNU General Public License as published by the Free Software
15 Foundation; either version 2 of the License, or (at your option) any later
16 version.
17
18 libDAI is distributed in the hope that it will be useful, but WITHOUT ANY
19 WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A
20 PARTICULAR PURPOSE. See the GNU General Public License for more details.
21
22 You should have received a copy of the GNU General Public License
23 along with libDAI in the file COPYING. If not, see http://www.gnu.org/licenses/
24
25 -------------------------------------------------------------------------------
26
27 Citing libDAI
28
29 If you write a scientific paper describing research that made substantive use
30 of this program, please cite the software appropriately, by mentioning the
31 fashion in which this software was used, including the version number.
32
33 An appropriate citation would be:
34
35 Joris M. Mooij et al. (2010) "libDAI 0.2.5: A free/open source C++ library for
36 Discrete Approximate Inference", http://www.libdai.org
37
38 or in BiBTeX format:
39
40 @misc{mooij2010libdai,
41 author = "Joris M. Mooij et al.",
42 title = "lib{DAI} 0.2.5: A free/open source {C}++ library for {D}iscrete {A}pproximate {I}nference",
43 howpublished = "http://www.libdai.org/",
44 year = 2010
45 }
46
47
48 Moreover, as a personal note, I would appreciate it to be informed about any
49 publications using libDAI at joris dot mooij at libdai dot org.
50
51 -------------------------------------------------------------------------------
52
53 About libDAI
54
55 libDAI is a free/open source C++ library (licensed under GPL 2+) that provides
56 implementations of various (approximate) inference methods for discrete
57 graphical models. libDAI supports arbitrary factor graphs with discrete
58 variables; this includes discrete Markov Random Fields and Bayesian Networks.
59
60 The library is targeted at researchers. To be able to use the library, a good
61 understanding of graphical models is needed.
62
63 The best way to use libDAI is by writing C++ code that invokes the library; in
64 addition, part of the functionality is accessibly by using the
65
66 * command line interface
67 * (limited) MatLab interface
68 * (experimental) python interface
69 * (experimental) octave interface.
70
71 libDAI can be used to implement novel (approximate) inference algorithms and to
72 easily compare the accuracy and performance with existing algorithms that have
73 been implemented already.
74
75 Features
76
77 Currently, libDAI supports the following (approximate) inference methods:
78
79 * Exact inference by brute force enumeration;
80 * Exact inference by junction-tree methods;
81 * Mean Field;
82 * Loopy Belief Propagation [KFL01];
83 * Fractional Belief Propagation [WiH03];
84 * Tree-Reweighted Belief Propagation [WJW03];
85 * Tree Expectation Propagation [MiQ04];
86 * Generalized Belief Propagation [YFW05];
87 * Double-loop GBP [HAK03];
88 * Various variants of Loop Corrected Belief Propagation [MoK07, MoR05];
89 * Gibbs sampler;
90 * Conditioned Belief Propagation [EaG09].
91
92 These inference methods can be used to calculate partition sums, marginals over
93 subsets of variables, and MAP states (the joint state of variables that has
94 maximum probability).
95
96 In addition, libDAI supports parameter learning of conditional probability
97 tables by Expectation Maximization.
98
99 Limitations
100
101 libDAI is not intended to be a complete package for approximate inference.
102 Instead, it should be considered as an "inference engine", providing various
103 inference methods. In particular, it contains no GUI, currently only supports
104 its own file format for input and output (although support for standard file
105 formats may be added later), and provides very limited visualization
106 functionalities. The only learning method supported currently is Expectation
107 Maximization (or Maximum Likelihood if no data is missing) for learning factor
108 parameters.
109
110 Rationale
111
112 In my opinion, the lack of open source "reference" implementations hampers
113 progress in research on approximate inference. Methods differ widely in terms
114 of quality and performance characteristics, which also depend in different ways
115 on various properties of the graphical models. Finding the best approximate
116 inference method for a particular application therefore often requires
117 empirical comparisons. However, implementing and debugging these methods takes
118 a lot of time which could otherwise be spent on research. I hope that this code
119 will aid researchers to be able to easily compare various (existing as well as
120 new) approximate inference methods, in this way accelerating research and
121 stimulating real-world applications of approximate inference.
122
123 Language
124
125 Because libDAI is implemented in C++, it is very fast compared with
126 implementations in MatLab (a factor 1000 faster is not uncommon). libDAI does
127 provide a (limited) MatLab interface for easy integration with MatLab. It also
128 provides a command line interface and experimental python and octave interfaces
129 (thanks to Patrick Pletscher).
130
131 Compatibility
132
133 The code has been developed under Debian GNU/Linux with the GCC compiler suite.
134 libDAI compiles successfully with g++ versions 3.4 up to 4.4.
135
136 libDAI has also been successfully compiled with MS Visual Studio 2008 under
137 Windows (but not all build targets are supported yet) and with Cygwin under
138 Windows.
139
140 Finally, libDAI has been compiled successfully on MacOS X.
141
142 Downloading libDAI
143
144 The libDAI sources and documentation can be downloaded from the libDAI website:
145 http://www.libdai.org.
146
147 Mailing list
148
149 The Google group "libDAI" (http://groups.google.com/group/libdai) can be used
150 for getting support and discussing development issues.
151
152 -------------------------------------------------------------------------------
153
154 Building libDAI under UNIX variants (Linux / Cygwin / Mac OS X)
155
156 Preparations
157
158 You need:
159
160 * a recent version of gcc (at least version 3.4)
161 * GNU make
162 * recent boost C++ libraries (at least version 1.37; however, version 1.37
163 shipped with Ubuntu 9.04 is known not to work)
164 * doxygen (only for building the documentation)
165 * graphviz (only for using some of the libDAI command line utilities)
166 * CImg library (only for building the image segmentation example)
167
168 On Debian/Ubuntu, you can easily install the required packages with a single
169 command:
170
171 apt-get install g++ make doxygen graphviz libboost-dev libboost-graph-dev libboost-program-options-dev libboost-test-dev cimg-dev
172
173 (root permissions needed).
174
175 On Mac OS X (10.4 is known to work), these packages can be installed easily via
176 MacPorts. If MacPorts is not already installed, install it according to the
177 instructions at http://www.macports.org/. Then, a simple
178
179 sudo port install gmake boost doxygen graphviz
180
181 should be enough to install everything that is needed.
182
183 On Cygwin, the prebuilt Cygwin package boost-1.33.1-x is known not to work. You
184 can however obtain the latest boost version (you need at least 1.37.0) from
185 http://www.boost.org/ and build it as described in the next subsection.
186
187 Building boost under Cygwin
188
189 * Download the latest boost libraries from http://www.boost.org
190 * Build the required boost libraries using:
191
192 ./bootstrap.sh --with-libraries=program_options,math,graph,test --prefix=/boost_root/
193 ./bjam
194
195 * In order to use dynamic linking, the boost .dll's should be somewhere in
196 the path. This can be achieved by a command like:
197
198 export PATH=$PATH:/boost_root/stage/lib
199
200 Building libDAI
201
202 To build the libDAI source, first copy a template Makefile.* to Makefile.conf
203 (for example, copy Makefile.LINUX to Makefile.conf if you use GNU/Linux). Then,
204 edit the Makefile.conf template to adapt it to your local setup. Especially
205 directories may differ from system to system. Platform independent build
206 options can be set in Makefile.ALL. Finally, run
207
208 make
209
210 The build includes a regression test, which may take a while to complete.
211
212 If the build is successful, you can test the example program:
213
214 examples/example tests/alarm.fg
215
216 or the more extensive test program:
217
218 tests/testdai --aliases tests/aliases.conf --filename tests/alarm.fg --methods JTREE_HUGIN BP_SEQMAX
219
220 -------------------------------------------------------------------------------
221
222 Building libDAI under Windows
223
224 Preparations
225
226 You need:
227
228 * A recent version of MicroSoft Visual Studio (2008 is known to work)
229 * recent boost C++ libraries (version 1.37 or higher)
230 * GNU make (can be obtained from http://gnuwin32.sourceforge.net)
231 * CImg library (only for building the image segmentation example)
232
233 For the regression test, you need:
234
235 * GNU diff, GNU sed (can be obtained from http://gnuwin32.sourceforge.net)
236
237 Building boost under Windows
238
239 Because building boost under Windows is tricky, I provide some guidance here.
240
241 * Download the boost zip file from http://www.boost.org/users/download and
242 unpack it somewhere.
243 * Download the bjam executable from http://www.boost.org/users/download and
244 unpack it somewhere else.
245 * Download Boost.Build (v2) from http://www.boost.org/docs/tools/build/
246 index.html and unpack it yet somewhere else.
247 * Edit the file boost-build.jam in the main boost directory to change the
248 BOOST_BUILD directory to the place where you put Boost.Build (use UNIX /
249 instead of Windows \ in pathnames).
250 * Copy the bjam.exe executable into the main boost directory. Now if you
251 issue "bjam --version" you should get a version and no errors. Issueing
252 "bjam --show-libraries" will show the libraries that will be built.
253 * The following command builds the boost libraries that are relevant for
254 libDAI:
255
256 bjam --with-graph --with-math --with-program_options --with-test link=static runtime-link=shared
257
258 Building libDAI
259
260 To build the source, copy Makefile.WINDOWS to Makefile.conf. Then, edit
261 Makefile.conf to adapt it to your local setup. Platform independent build
262 options can be set in Makefile.ALL. Finally, run (from the command line)
263
264 make
265
266 The build includes a regression test, which may take a while to complete.
267
268 If the build is successful, you can test the example program:
269
270 examples\example tests\alarm.fg
271
272 or the more extensive test program:
273
274 tests\testdai --aliases tests\aliases.conf --filename tests\alarm.fg --methods JTREE_HUGIN BP_SEQMAX
275
276 -------------------------------------------------------------------------------
277
278 Building the libDAI MatLab interface
279
280 You need:
281
282 * MatLab
283 * The platform-dependent requirements described above
284
285 First, you need to build the libDAI source as described above for your
286 platform. By default, the MatLab interface is disabled, so before compiling the
287 source, you have to enable it in Makefile.ALL by setting
288
289 WITH_MATLAB=true
290
291 Also, you have to configure the MatLab-specific parts of Makefile.conf to match
292 your system (in particular, the Makefile variables ME, MATLABDIR and MEX). The
293 MEX file extension depends on your platform; for a 64-bit linux x86_64 system
294 this would be "ME=.mexa64", for a 32-bit linux x86 system "ME=.mexglx". If you
295 are unsure about your MEX file extension: it needs to be the same as what the
296 MatLab command "mexext" returns. The required MEX files are built by issuing
297
298 make
299
300 from the command line. The MatLab interface is much less powerful than using
301 libDAI from C++. There are two reasons for this: (i) it is boring to write MEX
302 files; (ii) the large performance penalty paid when large data structures (like
303 factor graphs) have to be converted between their native C++ data structure to
304 something that MatLab understands.
305
306 A simple example of how to use the MatLab interface is the following (entered
307 at the MatLab prompt), which performs exact inference by the junction tree
308 algorithm and approximate inference by belief propagation on the ALARM network:
309
310 cd path_to_libdai/matlab
311 [psi] = dai_readfg ('../tests/alarm.fg');
312 [logZ,q,md,qv,qf] = dai (psi, 'JTREE', '[updates=HUGIN,verbose=0]')
313 [logZ,q,md,qv,qf] = dai (psi, 'BP', '[updates=SEQMAX,tol=1e-9,maxiter=10000,logdomain=0]')
314
315 where "path_to_libdai" has to be replaced with the directory in which libDAI
316 was installed. For other algorithms and some default parameters, see the file
317 tests/aliases.conf.
318
319 -------------------------------------------------------------------------------
320
321 Building the documentation
322
323 Install doxygen, graphviz and a TeX distribution and use
324
325 make doc
326
327 to build the documentation. If the documentation is not clear enough, feel free
328 to send me an email (or even better, to improve the documentation and send a
329 patch!). The documentation can also be browsed online at http://www.libdai.org.