Updated ChangeLog
[libdai.git] / ChangeLog
1 * Improved Makefile (libDAI now also builds out-of-the-box on MacOSX,
2 thanks to Dan Preston; merged Makefile.win and Makefile.shared into Makefile)
3 * Fixed bug in calcMarginal, calcPairBeliefs, calcPairBeliefsNew where
4 states with zero probability mass were clamped, leading to NaNs or assertion
5 errors (thanks to Dan Preston for reporting this)
6 * toc() now returns POSIX system time with maximum accuracy of microseconds
7 * Exception objects now remember their error code
8 * Added examples/example_springler.cpp
9
10
11 git 065eae35cbfcc36f1a945ae3053c80c23f366306
12 --------------------------------------------
13 * [Peter Gober] Made libDAI build out-of-the-box on Cygwin.
14 * [Frederik Eaton] Added Gibbs sampler to algorithms
15 * Improved documentation of include/dai/{bipgraph.h, prob.h, smallset.h,
16 var.h, varset.h, factor.h, enum.h} and of examples/example.cpp
17 Merged TODO and FILEFORMAT into doxygen documentation
18 * examples/
19 - Moved example.cpp to examples/
20 - Added examples/example_bipgraph.cpp
21 - Added examples/example_varset.cpp
22 * Interface changes:
23 - Replaced TProb<T>::log0() by TProb<T>::log(true)
24 - Replaced TProb<T>::takeLog0() by TProb<T>::takeLog(true)
25 - Replaced TFactor<T>::log0() by TFactor<T>::log(true)
26 - Removed TProb<T>::hasNonPositives()
27 - Renamed smallSet<T> to SmallSet<T>
28 - Removed TFactor<T>::divided_by, use operator/ instead
29 - Removed TFactor<T>::divide, use operator/= instead
30 - Removed TFactor<T>::partSum(const VarSet&), use marginal(const VarSet&,true) instead
31 - Improved constructors of TProb and TFactor to use iterators instead of pointers
32 * Miscellaneous small changes and cleanups:
33 - Changed regression test so that it also works under Windows
34 - Changed output stream operator<< for Var and VarSet
35 - Added TProb::draw() function, which draws a random index
36 - Cleanup of matlab interface code
37 - Small improvement of utils/fginfo
38 - Small cleanup of BP code
39 - Switched Makefile.win to GNU Make syntax
40
41
42 libDAI-0.2.2 (2008-09-30)
43 -------------------------
44
45 New features:
46 * Approximate inference methods now report the number of iterations needed.
47 * Added damping to various algorithms to improve convergence properties.
48 * Added more features to utils/createfg for creating factor graphs.
49 * Added ExactInf class for brute force exact inference.
50 * [Giuseppe Pasino] Added "logdomain" property to BP, a boolean that controls
51 whether calculations are done in the log-domain or in the linear domain;
52 doing calculations in the log-domain may help if the numerical range
53 of a double is too small.
54 * [Claudio Lima] Added Max-Product functionality to BP.
55 * Improved documentation.
56
57 Improved architecture:
58 * Added Exceptions framework.
59 * Pervasive change of BipartiteGraph implementation (based on an idea by
60 Giuseppe Passino). BipartiteGraph no longer stores the node properties
61 (former _V1 and _V2), nor does it store a dense adjacency matrix anymore,
62 nor an edge list. Instead, it stores the graph structure as lists of
63 neighboring nodes. This yields a significant memory/speed improvement for
64 large factor graphs, and is more elegant as well. Iterating over neighbors is
65 made easy by using boost::foreach.
66 * Added conditional compilation of inference methods.
67 * VarSet is now implemented using a std::vector<Var> instead of a
68 std::set<Var>, which yields a significant speed improvement. Furthermore,
69 the implementation has been generalized, resulting in the small_set<T> class
70 which can be used to represent sets of small cardinality; VarSet is the
71 specialization with T = Var.
72 * Improved ClusterGraph implementation, yielding significant speedups
73 for the JunctionTree algorithm on large factorgraphs.
74
75 Code cleanup:
76 * Moved everything into namespace "dai".
77 * Renamed DEBUG to DAI_DEBUG to avoid conflicts.
78 * Replaced ENUM2,ENUM3,ENUM4,ENUM5,ENUM6 by single DAI_ENUM macro.
79 * Removed utils/remove_short_loops and matlab/remove_short_loops.
80 * Replaced sub_nb class in mr.h by boost::dynamic_bitset.
81 * Improved index.h:
82 - Renamed Index -> IndexFor
83 - Added some .reserve()'s to IndexFor methods which yields a
84 25% speedup of testregression
85 - Replaced multind by Permute
86 - Added MultiFor
87 - Added State
88 * New funcstionality of factor.h.
89 * Moved Properties and MaxDiff frameworks from InfAlg to each individual
90 inference algorithm, because the Properties framework was not as
91 convenient as I hoped, and not every inference algorithm needs a maxdiff
92 variable. Also, replaced some FactorGraph functionality in InfAlg by a
93 function that returns the FactorGraph. The result is cleaner (less
94 entangled) code.
95 * Removed x2x.
96 * Replaced Complex with real numbers (negative potentials are just too rare
97 to warrant the additional "complexity" :)).
98
99 Miscellaneous improvements:
100 * Now compiles also with MS Visual C++ (thanks to Jiuxiang Hu) and with
101 GCC under cygwin.
102 * Contributions by Peter Gober:
103 - Renamed variable _N in mr.* for compatibility with g++ under cygwin.
104 * Misc contributions by Giuseppe Passino:
105 - removed "using namespace std;" from header files - bad practice;
106 - moved header files in include/dai and sources in src;
107 - changed #ifndefs to GNU style;
108 - added extra warning checks (-W -Wextra) and fixed resulting warnings;
109 - dai::TProb:
110 o removed copy constructor and assignment operators (redundant);
111 o implementation of some methods via STL algorithms;
112 o added methods takeExp, takeLog, takeLog0 for transformation in-place;
113 o explicit constructor (prevents implicit conversion from size_t to TProb);
114 o added operator+,+=,-,-=, with argument T (for calculations in log-scale);
115 * Misc contributions by Christian Wojek:
116 - New FactorGraph constructor that constructs from given ranges of factors
117 and variables;
118 - Optimization of FactorGraph constructors using tr1::unordered_map.
119 * FactorGraph constructors no longer check for short loops (huge speed
120 increase for large factor graphs), nor for negative entries. Also, the
121 normtype is now Prob::NORMPROB by default.
122 * Improved MaxSpanningTreePrims algorithm (uses boost::graph).
123
124 Interface changes:
125 * VarSet::
126 - VarSet::stateSpace() -> nrStates(const VarSet &)
127 - VarSet( const std::set<Var> ) -> VarSet( begin, end, sizeHint=0 )
128 - VarSet( const std::vector<Var> ) -> VarSet( begin, end, sizeHint=0 )
129 - removed bool operator||
130 - operator&&(const VarSet&) -> intersects(const VarSet&)
131 - operator&&(const Var&) -> contains(const Var&)
132 * FactorGraph::
133 - delta(const Var &) -> delta(size_t)
134 - Delta(const Var &) -> Delta(size_t)
135 - makeCavity(const Var &) -> makeCavity(size_t)
136 - vars() -> vars
137 - factors() -> factors
138 - removed MakeFactorCavity(size_t)
139 - removed ExactMarginal(const VarSet &)
140 - removed ExactlogZ()
141 - removed updatedFactor(size_t)
142 - removed _normtype and NormType()
143 - removed hasShortLoops(...) and removeShortLoops(...)
144 - WriteToDotFile(const char *filename) -> printDot( std::ostream& os )
145 - undoProb(size_t) -> restoreFactor(size_t)
146 - saveProb(size_t) -> backupFactor(size_t)
147 - undoProbs(const VarSet &) -> restoreFactors(const VarSet &)
148 - saveProbs(const VarSet &) -> backupFactors(const VarSet &)
149 - ReadFromFile(const char*) returns void (throws on error)
150 - WriteToFile(const char*) returns void (throws on error)
151 - removed hasNegatives()
152 * RegionGraph::
153 - nr_ORs() -> nrORs()
154 - nr_IRs() -> nrIRs()
155 - ORs() -> ORs
156 - IRs() -> IRs
157 * *::Regenerate() -> *::construct()
158 * Renamed Index -> IndexFor
159 * Diffs:
160 - max() -> maxDiff()
161 - max_size() -> maxSize()
162 * Prob::max() -> Prob::maxVal()
163 * Factor::
164 - max() -> maxVal()
165 - part_sum() -> partSum()
166 * toc() in util.h now returns seconds as a double
167 * VarSet::operator&&
168 * Properties -> PropertySet
169
170
171 libDAI-0.2.1 (2008-05-26)
172 -------------------------
173
174 Bugfix release.
175 * added missing cstdio header in util.h
176 * fixed Properties in MR_CLAMPING_* and MR_EXACT_*
177 * added description of the factor graph fileformat
178 * improved Makefile
179
180
181 libDAI-0.2.0 (2006-11-30)
182 -------------------------
183
184 First public release.
185
186
187 0.1.5 (2006-11-30)
188 ------------------
189
190 Regressions
191
192 - tests/testlcbp and tests/testlcbp are broken.
193 - EXACT method does not work anymore.
194 - The Properties framework gives a speed penalty because of the lookup
195 costs involved; inner loops should be optimized.
196
197 General framework
198
199 - DAIAlg is now a template class; typedefs for DAIAlg<FactorGraph> and for
200 DAIAlg<RegionGraph> are provided. In this way, we do not have to write "wrapper"
201 functions to forward functionality from either FactorGraph or RegionGraph
202 to DAIAlg. Functionality like clamping can be implemented in FactorGraph
203 and in RegionGraph and no explicit interface is needed in descendants.
204 - New abstract base class InfAlg added, representing an inference algorithm,
205 from which DAIAlg<T> inherits. This solves the incompatibility problems of
206 DAIAlg<T> for different T (e.g. DAIAlg<FactorGraph> was incompatible with
207 DAIAlg<RegionGraph>). More work is required to reduce code duplication
208 (make FactorGraph part of InfAlg).
209 - Added generic interface (nrVars(), Vars(), nrFactors(), factor(size_t),
210 beliefs(), belief(VarSet &), ...) to InfAlg and descendants.
211 - Added a saveProbs/undoProbs interface to InfAlg and descendants that enables
212 one to save a few factors, modify them (e.g. clamp them), and then restore them
213 to their old values. Undo should also init the corresponding messages / beliefs.
214 This can be used if a given factor graph repeatedly needs to be clamped in
215 different ways and an approximation method is run for each clamping; using the
216 saveProbs/undoProbs can give a significant speed increase.
217 - Switched to a general Properties framework that handles the parameters of
218 all inference methods in a uniform manner. The Properties class is a map of
219 several properties in boost::any objects, indexed by their names (strings).
220 It can read from a stream and write to a stream. It is recursive, in the sense
221 that a Properties object can hold a variable of type Properties as well.
222 - Added a generic way of constructing inference algorithms from a factor graph,
223 name and properties object. Added the newInfAlg function which constructs
224 the requested object. This is used by LCBP, the Matlab interface and the
225 command line (test) interface.
226 - Added a generic enum framework for enum parameters. Although implemented as a
227 hack, it is quite useful in that it drastically simplifies and reduces the
228 amount of code for handling enum parameters.
229 - Provided generic functions for calculating marginals in different ways that
230 work for all approximate inference methods.
231
232 Bugfixes
233
234 - Fixed GBP free energy.
235 - Fixed bug in junctiontree (it didn't init the _vars variable).
236 - Corrected two bugs in operator&& and operator|| in VarSet (they returned
237 the logical NOT of what they should return).
238 - Fixed bug in RegionGraph::RecomputeOR(s).
239 - Fixed bug in utils/create_dreg_fg:
240 graph structure was not random for given parameters (forgot to call srand()).
241 - TreeEP bug workaround: use the complete junction tree instead of a subtree.
242 - Fixed bug in JTree::HUGIN() and JTree:ShaferShenoy() in case of junction tree
243 that consists of one outer region only.
244 - Fixed INIT bug in LCBP2::UpdatePancake().
245 - Fixed MaxDiffs flow (except for MR).
246
247 New functionality
248
249 - HAK supports several default cluster choices:
250 minimal (only factors)
251 delta (Markov blankets)
252 loop (all loops consisting of loops consisting of <loopdepth> or less variables)
253 Only the maximal clusters are used as outer clusters.
254 - Implemented TreeEP. It generalizes the heuristic method described in the
255 Minka & Qi paper for obtaining a tree with the most relevant interactions to
256 higher order interactions. Almost all optimizations described in the Minka & Qi
257 paper are used, except that evidence is passed over the whole tree instead of
258 relevant subsets (the latter is almost implemented but buggy). Also added
259 alternative (worst-case) algorithm that uses a maximum spanning tree on the
260 weighted graph where the weight between neighbours i and j is given by
261 N(psi,i,j), where psi is the product of all the factors involving both i and j
262 (which is an upper bound on the effective interaction between i and j).
263 - Implemented MR (MontanariRizzo) based on Bastian's code, but extended it
264 to be able to handle connectivities larger than 3 (in principle, up to 32).
265 It supports different initialization methods (the original RESPPROP,
266 the CLAMPING method and EXACT which uses JTree) and different update methods
267 (FULL and LINEAR).
268 - Implemented LCBP2, an analogon of LCBP which represents pancakes as little
269 networks and uses some approximate inference method on them for calculating
270 marginals.
271 - Now there are several LCBP variants (LCBP, LCBPI, LCBPJ, LCBPK, LCBPL);
272 LCBPJ works only for pairwise, LCBPK is LCBP improved for higher order
273 interactions and LCBPL is LCBPI improved for higher-order interactions.
274 - Wrote one single program utils/createfg for creating various types of
275 random factor graphs.
276 - Wrote utility to visualize factor graphs using graphviz.
277 (it uses the BOOST Program Options library)
278 - Added fginfo utility that displays some info about a .fg file.
279 - Implemented Factor::strength function that calculates the potential strength
280 N(psi,i,j) between variables i and j as described in cs.IT:0504030
281 - Wrote a general MatLab interface matlab/dai (similar to tests/test);
282 this unified the matlab functions dai, dai_bp, dai_mf, dai_jt, dai_tep, dai_cvm.
283 - Added MATLAB routine that returns contraction matrix A for BP convergence analysis.
284 - Implemented a MATLAB interface ai_potstrength for Factor::strength
285 - Added Martijn's x2x
286
287 Improvements of existing code
288
289 - Reimplemented RegionGraph and descendants: a RegionGraph ISA FactorGraph
290 and also a BipartiteGraph<FRegion,Region>. It now also keeps a map that
291 associates outer region indices to factor indices (no powers yet, this
292 is deemed superfluous) and provides functions to recompute (some of) the
293 outer regions from the factors.
294 - InfAlg descendants run() methods now stop immediately and return NAN in case
295 they detect NANs. Only BP does not do NAN checking for performance reasons.
296 - LCBP now works with factors containing zeroes (by defining x/0 = 0).
297 - HAK, GBP and DoubleLoop now conform to the standards for verbose reporting,
298 timing and convergence criteria.
299 - Implemented logZ() for JTree. It does the calculation during message-passing.
300 - Marginal2ndO now optionally divides by the single node beliefs (to the power n-2);
301 hopefully this will give more accurate approximations.
302 - Marginal and Marginal2ndO (optionally) use the new saveProbs/undoProbs functionality
303 for a faster way of calculating marginals, which does not require a call to init()
304 nor cloning the whole object for each clamping. This leads to a significant speedup.
305 - LCBP (and LCBP2) now have complete flexibility in the specification of the
306 inner method, i.e. the method used to generate the initial cavity approximations.
307 One can pass two strings, a name and a properties string, and LCBP simply invokes
308 newInfAlg to construct the corresponding inference algorithm and uses the generic
309 marginal functions to approximate cavity marginals.
310 - Replaced the global "method" variable by local properties and removed ai.h
311 - Added some methods to Factor (operators *, *=, /, /= with doubles as
312 second argument, operators -, +=, -= with other Factors as second
313 arguments, randomize(), RemoveFirstOrderInteractions) and similar
314 operations to Prob
315 - Moving towards boost::program_options for handling command line arguments
316 (tests/test is done).
317 - Renamed some FactorGraph methods:
318 nr_vars -> nrVars
319 nr_factors -> nrFactors
320 varind -> findVar
321 factorind -> findFactor
322 makeFacCavity -> makeFactorCavity
323 - LCBP_SEQMAXRES has been removed because it did strange things.
324 - Implemented RandomDRegularGraph
325 - Implemented JTree::calcMarginal for marginals that are not confined
326 within one cluster (using cut-set conditioning).
327 - Added isConnected() method to FactorGraph (some methods do not work with
328 disconnected factor graphs).
329 - Pair beliefs are now calculated in a symmetrical way by calcPairBeliefs
330 - Removed single node interaction "correction" code from clamping methods
331 - Removed calcCavityDist and calcCavityDist2ndO
332 - No longer depends on GSL.
333 - Increased portability by combining platform dependant utility functions
334 in util.{h,cpp}.
335 - Wrote *.m files providing help
336
337 Testing framework
338
339 - Made a new and significantly improved testing framework that provides most
340 functionality from the command line.
341 - The basis is provided by tests/test, which uses the newInfAlg functionality
342 and enables the user to easily compare from the command line different
343 inference methods on a given factorgraph. All parameters can be specified.
344 Output consists of CPU time, average and maximum single variable marginal
345 errors, relative logZ error and MaxDiff().
346 - tests/aliases.conf contains aliases for standard combinations of methods
347 and their options (which can be used in tests/test).
348 - tests/large contains several bash/python scripts that create random factor
349 graphs, compare several approximate inference algorithms (using tests/test) and
350 allow for easy visualization of the results using PyX.
351 - Added several .fg files for test purposes to /tests (e.g. two ALARM versions
352 alarm.fg and alarm_bnt.fg; testfast.fg, a 4x4 periodic Ising grid for
353 regression testing).
354 - Added a regression test to the Makefile which is included in the standard
355 target. It compares all inference methods on tests/testfast.fg with the
356 results stored in tests/testfast.out
357
358 Miscellaneous
359
360 - Expanded all tabs to spaces (":set tabstop 4\n:set expandtab\n:retab" in vim)
361 - Experimental MATLAB code added for StarEP approximation on cavity
362 - Renamed project to libDAI and changed directory name accordingly.
363 - Renamed JunctionTree to JTree.
364 - Fixed licensing (now it's officially GPL).
365 - Improved README
366
367
368 revision 252
369 ------------
370
371 Functionality
372 - Added RegionGraph, GBP, CVM and HAK (double-loop).
373 - Added JunctionTree (with two update algorithms, HUGIN and Shafer-Shenoy), which is a
374 RegionGraph.
375 - NormType is now chosen automatically (in case of negative factors, Prob::NORMLINF is used,
376 otherwise the default Prob::NORMPROB is used). Also, in case of negative factors, the
377 RegionGraph constructors assign each Factor to a unique outer region instead of dividing
378 it over all subsuming outer regions. See README for when negative factors are known to work
379 and when not.
380 - FactorGraph::FactorGraph(const vector<Factor>) only gives a warning in case of short loops,
381 it does not automatically merge factors anymore.
382 - Removed BP_SEQMAXRESNOCLEAR (all cavity initialization methods now are implicitly NOCLEAR)
383 - Added MATLAB interface functions ai_readfg, ai_removeshortloops and ai_bp
384 - Added LCBP-III type that should be equivalent to LCBP-II, but can handle zeroes
385 in potentials. Note that it is significantly slower than LCBP-II (and has to be reimplemented
386 such that it does not store the complete pancakes, but represents them as little factor graphs).
387
388 Implementation / code
389 - Made a seperate type WeightedGraph, which until now only implements Prim's
390 maximal spanning tree algorithm and is only used by the junction tree code. It might
391 be good to make it a class somewhere in the future and extend it's interface.
392 - Made a seperate class ClusterGraph, which is only used by the junction tree
393 code. It's main purpose is a graph-theoretical variable elimination algorithm.
394 - Implemented the heuristic "minimum-new-edges-in-clique-graph" for variable elimination.
395 - Massive code cleanup, moving towards "generic" programming style, using
396 multiple inheritance and polymorphism.
397 o BP, LCBP, MF, HAK and JunctionTree now inherit from a common DAIAlg class
398 o Made generic functions Marginal, Marginal2ndO, calcCavityDist, calcCavityDist2ndO, clamp
399 that can be used by FactorGraph-based DAIAlgs.
400 o Created TProb<T> class, which stores a probability vector (without the accompanying indexing
401 and VarSet) and provides functionality for it (which is used by TFactor<T>).
402 o Rewrote the VarSet class. It now caches its statespace(). It now privately inherits from set<Var>.
403 I had to overload the insert methods of set<Var> so that they calculate the new statespace.
404 o Rewrote the TFactor class. The TFactor class now HAS a TProb and HAS a VarSet.
405 - Rewrote BP to use the new TProb<T> interface. Performance of BP improved up to a factor 6 by:
406 o using Prob's instead of Factor's;
407 o splitting the multiplication of the messages into two parts (thanks to Vicenc!);
408 o optimizing the calculation of the beliefs (only the message calculations were optimized till now).
409 o replacing FactorGraph::_nb1 and _nb2 (which were set<size_t>) by vector<size_t>
410 - LCBP now seperately stores cavitydists and pancakes. Added InitPancakes() method
411 that takes the cavitydists and multiplies them with the relevant factors. This
412 resulted in an API change in AI which now accepts and returns initial cavitydists
413 instead of initial pancakes.
414
415 Minor changes
416 - Started writing DoxyGen documentation
417 - Renamed lcptab2fg.m matlab/ai_writefg.m
418 - Moved all matlab stuff to matlab/
419 - More detailed reporting (also reports clocks used).
420 - Marginal and Marginal2ndO now use *differences* in logZ to avoid NaNs.
421 - Improved testing suite.
422 - Removed logreal support.
423 - FactorGraph now also supports input streams and ignores comment lines in .fg files.
424 - Added tests/create_full_fg.cpp and tests/create_ising_fg.cpp which create
425 full and periodic 2D Ising networks according to some command line parameters.
426 - Now logZ really returns logZ instead of -logZ.
427 - Added FactorGraph::WriteToDotFile
428
429
430 0.1.4 (2006-04-13)
431 ------------------
432 - Added file IO routines to read and write factorgraphs.
433 - Added L-infinity normalization for use with (partially) negative factors.
434 - Renamed BetheF, MFF to logZ, which now use complex numbers to be able to
435 handle (partially) negative factors.
436 - Added test suite.
437 - All probabilities are now represented using double instead of LogReal<double>.
438 - Combined Alg and Alg3 into LCBP. Several update schemes possible.
439 - Combined several variants of BP into doBP. Several update schemes possible.
440 Now uses precalculated indices for optimization.
441 - Renamed Node -> Var and Potential -> Factor.
442 - Extensive code cleanup. More use of OO techniques. Changed API.
443 - MaxIter now means maximum number of passes (corresponding to number of
444 _parallel_ updates).
445 - MaxDiff now means the maximum L-infinity distance between the updated and
446 original single variable beliefs, for all AI methods.
447 - Implemented RBP which converges faster than BP for difficult problems.
448 - Now uses method parameter which is a bitmask combining outmethod and inmethod
449 (see ai.h).
450
451
452 0.1.3 (2006-03-22)
453 --------------------
454 - All AI methods now return maxdiff
455 - ai.cpp:
456 o Now returns maxdiffinner and maxdiffouter
457 o New BP2ndO innermethod (estimate only 2nd order cavity interactions)
458 o New InitCav outermethod (only create initial cavity distributions)
459 - bp.cpp:
460 o New CavityDist2ndO which estimates 2nd order cavity interactions
461 - Makefile:
462 o Bugfix: removed dependencies on algwim.*
463
464
465 0.1.2 (2006-02-28)
466 --------------------
467 - Cleaned up alg.cpp (removed Alg2 and its corresponding data structures).
468 - Added the possibility to provide initial cavity distributions as an input
469 argument to ai (not much error checking is done, so be careful).
470 - Potentials2mx now correctly sets the dimensions of the P field (i.e. for
471 the output arguments Q, Q0 of ai).
472 - Removed algwim.* since it does not work.
473
474
475 0.1.1 (2006-02-28)
476 --------------------
477 - The constructors of (Log)FactorGraph and LogFactorGraph from a
478 vector<(Log)Potential> now merge potentials to prevent short loops (of length
479 4) in the factor graph. These are used in ai to construct the factor graphs
480 from the psi argument. If compiled with DEBUG defined, the method calc_nb()
481 of BipGraph checks for the existence of short loops.
482 - Changed calling syntax of ai (now the actual syntax *does* correspond to its
483 description in the help).
484 - ai does not hook cout anymore (which caused weird segfaults).
485 - Fixed a bug in an assert statement in the matlab interface code in ai.cpp.
486 - Removed network.* since it is not useful.
487
488
489 0.1.0 (2006-02-28)
490 --------------------
491 First version worthy a version number.