R-LAIR: Riverside Lab for Artificial Intelligence Research
Continuous Time Bayesian Network Reasoning and Learning Engine (CTBN-RLE)
v1.1.1: Install Instructions
How to compile:
===============
This code has been tested under Unix with the Gnu tools and Windows
with Visual C++. You may want to read the "Installing Graphviz" section
first if you plan to use the visualization portion of the libraries.
UNIX:
The Makefile in src/ will make ctbn.a (a library of all of the object files).
"make" in demo/ will make the demos (including making the objects in src/
if necessary), and "make" tst/ will make all of the unit/regression tests.
Running the python script tst/unit_test.py will run each of the test
scripts to confirm that the calculations are being performed correctly.
Windows:
Under Visual C++, you must create your own project. Add all of the source
files in src/ and all of the header files in hdr/. Adding one "main" file
(from tst/ or demo/) should allow the entire project to compile. You may
have a number of compile warnings about conversions from "size_t" to "int."
These may be safely ignored.
Using CMake:
============
Starting with release 1.1.0, we have added support for cmake.
To use it with make (on Unix), look at the build.sh script in the base
directory. It shows how to configure glpk and then run cmake followed by
make (from the directory ctbn_build_dir). All resulting executables will
be in the ctbn_build_dir under different directories.
We hope to improve cmake support in future releases.
Installing Graphviz:
====================
graphiviz is a graph layout program available from www.graphviz.org.
The CTBNRLE uses the "dot" executable from it. We have successfully tested
CTBNRLE with graphviz version 2.14.1 (Oct 20, 2007) and version 2.26.3 (Jan
26, 2010).
Install graphviz as per the instructions on their website. If the
executable is not in your path, or you would like to explicitly state
where to find it, you should edit src/graphic.cc and change the definition of
DOTCOMMAND to the string containing the executable (with path) to be executed
to run the "dot" program of graphviz.
Tests:
======
"make test" from the cmake build directory or "python ./unit_test.py"
from the tst/ directory will run a bank of tests. Note that the SEM test
will take a few minutes (much longer if compiler speed optimizations are
not on). Very occasionally one of the sampling tests will fail. Rerunning
(ie with a different random seed) will fix it.
Continuous Time Bayesian Network Reasoning and Learning Engine (CTBN-RLE)
v1.1.1: Install Instructions
How to compile: =============== This code has been tested under Unix with the Gnu tools and Windows with Visual C++. You may want to read the "Installing Graphviz" section first if you plan to use the visualization portion of the libraries. UNIX: The Makefile in src/ will make ctbn.a (a library of all of the object files). "make" in demo/ will make the demos (including making the objects in src/ if necessary), and "make" tst/ will make all of the unit/regression tests. Running the python script tst/unit_test.py will run each of the test scripts to confirm that the calculations are being performed correctly. Windows: Under Visual C++, you must create your own project. Add all of the source files in src/ and all of the header files in hdr/. Adding one "main" file (from tst/ or demo/) should allow the entire project to compile. You may have a number of compile warnings about conversions from "size_t" to "int." These may be safely ignored. Using CMake: ============ Starting with release 1.1.0, we have added support for cmake. To use it with make (on Unix), look at the build.sh script in the base directory. It shows how to configure glpk and then run cmake followed by make (from the directory ctbn_build_dir). All resulting executables will be in the ctbn_build_dir under different directories. We hope to improve cmake support in future releases. Installing Graphviz: ==================== graphiviz is a graph layout program available from www.graphviz.org. The CTBNRLE uses the "dot" executable from it. We have successfully tested CTBNRLE with graphviz version 2.14.1 (Oct 20, 2007) and version 2.26.3 (Jan 26, 2010). Install graphviz as per the instructions on their website. If the executable is not in your path, or you would like to explicitly state where to find it, you should edit src/graphic.cc and change the definition of DOTCOMMAND to the string containing the executable (with path) to be executed to run the "dot" program of graphviz. Tests: ====== "make test" from the cmake build directory or "python ./unit_test.py" from the tst/ directory will run a bank of tests. Note that the SEM test will take a few minutes (much longer if compiler speed optimizations are not on). Very occasionally one of the sampling tests will fail. Rerunning (ie with a different random seed) will fix it.