OpenPNL

OpenPNL copied to clipboard

PNL -- Probabilistic Networks Library. Release 1.0. 31-July-2006

Requirements Operating system: MS Windows 98/Me/2000/XP Linux Compiler: Visual C++ 6.0 (Intel Compiler 7.0 may used as compiler for a Visual Studio Environment) gcc 3.x.x, 4.x.x icc 8.x icc 9.0, 9.1

Directory tree.

    bin(*)             -- executable files and DLLs
    lib(*)             -- static libraries and stub libraries (for using DLLs)
    c_pgmtk            -- root folder for C++ version of PNL
       examples        -- example applications
          data         -- folder containing data files, which is used in examples
       include         -- include files for external interface
       make            -- project definition files 
       src             -- source code of library
          include      -- internal header files
       tests           -- set of tests
          include      -- internal header files for tests
          src          -- source code for tests
          make         -- project definition files for tests        
          testdata     -- data files used by tests 
          !readme!.txt -- read it before you start building tests
       high            -- high level API for PNL (experimental for now)
    cxcore             -- openCV core. Used for operation with sparse matrices
    doc                -- users guide and reference manual           
    trs                -- TRS test system
       include         -- header files
       make            -- project definition files
       src             -- source files

(*) The directory and its content are generated during the build process.

Building the library, examples and tests for C/C++ version from Developer Studio 6.0

To build the library and utilities from Developer Studio 6.0 do the following:

1. Start Microsoft Developer Studio 6.0.

2. Open workspace "c_pgmtk/make/pnl.dsw". It contains the following projects:

   Project... For...

   _build_all All components provided by workspace ex_param Example of using evidence class gibbs Example of using Gibbs inference sampling inf_learn_bnet Example of using inference and learning classes for BNets inf_learn_dbn Example of using inference and learning classes for DBNs learn_param Example of using learn class mixture_gaussian_bnet Example of mixture gaussian bnet creation pnl C++ version of PNL testLIMID Example of using LIMID inference for Influence Diagrams testParPNL Example of using parallel methods for some algorithms test_pnl_c Tests for C++ version of PGMTk testSL Test on structure learning of BNet trial Example of working with junction tree inference engine trs TRS test system use_matrix Example of operating with matricies

3. Build project _build_all to build library, examples and tests.

Notes: (a) Configurations "Win32 Debug" and "Win32 Release" build DLL version of the library, examples and tests that link this DLL.

(b) Debug variants of library, examples and tests have the suffix "d", for example: "pnld.dll", "triald.exe".

(c) Configurations "Win32 Parallel Debug" and "Win32 Parallel Release" build DLL version of the library, that contains parallel classes. MPI or(and) OpenMP versions can be built by using "BUILD_MPI" or(and) "BUILD_OMP" precompiler's definitions. OpenMP case suppose to use "/Qopenmp" key as a compiler's option.

Building the library, examples and tests for C/C++ version from Linux with gcc

1. Go to the root directory (it contain this file and changes.txt)
2. Run './configure.gcc'
3. Run 'make' to compile sources
4. Run 'make check' to compile and launch test suite (optionally)
5. Run 'make install' to install library

Notes:

• Step 2 (Run './configure.gcc') should be run on initial or on compiler changing
• If you want to install library to some directory instead of '/usr/local' (as default), you can use '--prefix' option of 'configure' script in 'configure.gcc' file (run './configure -h' to read more)
• You can use object directory to build library. In this case step 2 looks like 'SRCROOT/configure.gcc', where 'SRCROOT' is relative path to source root directory
• If you have some error during compiling or if you want to view compiling message later, run 'make 2>&1 | tee compiling.log' instead of 'make'

Building the library, examples and tests for C/C++ version from Linux with icc (Intel compiler)

1. Go to the root directory (it contain this file and changes.txt)
2. Run './configure.icc'
3. Run 'make' to compile sources
4. Run 'make check' to compile and launch test suite (optionally)
5. Run 'make install' to install library

Notes:

• Step 2 (Run './configure.icc') should be run on initial or on compiler changing
• If you want to install library to some directory instead of '/usr/local' (as default), you can use '--prefix' option of 'configure' script in 'configure.icc' file (run './configure -h' to read more)
• You can use object directory to build library. In this case step 2 looks like 'SRCROOT/configure.icc', where 'SRCROOT' is relative path to source root directory
• If you want to compile pnl with parallel functionality (OpenMP parallel mode of pnl) you have to define CXXFLAGS variable as '-openmp' and define BUILD_OMP in pnlParConfig.hpp as macro of preprocessor
• If you want to compile pnl with parallel functionality (Cluster OpemMP parallel mode of pnl) you have to define CXXFLAGS variable as '-cluster-openmp'
• If you have some error during compiling or if you want to view compiling message later, run 'make 2>&1 | tee compiling.log' instead of 'make'