(learning probabilistic and possibilistic graphical models)
|32 bit||64 bit||(32/64 bit only for executable)|
|lnxines.zip||(560 kb)||lnxines.zip||(649 kb)||GNU/Linux executables|
|winines.zip||(438 kb)||winines.zip||(486 kb)||Windows console executables|
|ines.zip||(285 kb)||ines.tar.gz||(238 kb)||C sources, package version 4.1 (2014.10.24)|
Note: The table package contains some auxiliary programs for preprocessing the data files.
INeS (Induction of Networks Structures) is a set of programs to learn a graphical model (Bayesian network or possibilistic network) from a dataset of sample cases, to generate a random dataset from a Bayesian network, to evaluate learned networks w.r.t. a test dataset and a reference network, and to measure the strengths of conditional dependences.
A brief description of how to apply these programs can be found in the file ines/ex/readme in the source package. The scripts djc_prob, djc_poss, and djc_local in the directory ines/djc may also be helpful.
If you have trouble executing the programs on Microsoft Windows, check whether you have the Microsoft Visual C++ Redistributable Packages for Visual Studio 2015 installed, as the library was compiled with Microsoft Visual Studio 2015.
The theory underlying this program is described in detail in the books:
Note: The learning results reported in these books were produced on a GNU/Linux system using the rand48 random number generator, which is not available on a Microsoft Windows System (but should be available on an Apple OSX system). In order to avoid different results on different systems, the above programs now use a different random number generator by default (see util/src/random.[ch]). However, on a GNU/Linux system, the results reported in the books can still be reproduced exactly by recompiling the programs with
make all ADDFLAGS="-DRAND_LIB48"
which compiles the programs in such a way that they use the rand48 random number generator.