BayesianNetworktool

所属分类:Java编程
开发工具:Java
文件大小:2650KB
下载次数:169
上传日期:2010-07-15 23:16:37
上 传 者luoyaping8555
说明:  一款非常优秀的开源贝叶斯网络工具包,java程序编写的。
(A very good open source Bayesian network toolkit, java programming in.)

文件列表:
bnj-贝叶斯网络java\bnj (513, 2003-03-18)
bnj-贝叶斯网络java\bnj.bat (465, 2003-03-18)
bnj-贝叶斯网络java\build (878, 2003-03-18)
bnj-贝叶斯网络java\build.bat (841, 2003-03-18)
bnj-贝叶斯网络java\etc\log4j.properties (770, 2003-03-18)
bnj-贝叶斯网络java\examples\alarm\alarm.bif (13460, 2003-03-18)
bnj-贝叶斯网络java\examples\alarm\alarm.dsc (13771, 2003-03-18)
bnj-贝叶斯网络java\examples\alarm\alarm.net (11447, 2003-03-18)
bnj-贝叶斯网络java\examples\alarm\alarm.xeb (713, 2003-03-18)
bnj-贝叶斯网络java\examples\alarm\alarm.xml (15825, 2003-03-18)
bnj-贝叶斯网络java\examples\alarm\alarmevidence.txt (615, 2003-03-18)
bnj-贝叶斯网络java\examples\alarm\alarmProbs.txt (1024, 2003-03-18)
bnj-贝叶斯网络java\examples\asia\asia-2.xml (3217, 2003-03-18)
bnj-贝叶斯网络java\examples\asia\asia-57.net (3108, 2003-03-18)
bnj-贝叶斯网络java\examples\asia\asia-new6.net (3237, 2003-03-18)
bnj-贝叶斯网络java\examples\asia\asia.net (1439, 2003-03-18)
bnj-贝叶斯网络java\examples\asia\asia.xeb (464, 2003-03-18)
bnj-贝叶斯网络java\examples\asia\asia.xml (3133, 2003-03-18)
bnj-贝叶斯网络java\examples\asia\asia1000data.arff (68043, 2003-03-18)
bnj-贝叶斯网络java\examples\asia\asia1000data.xml (74971, 2003-03-18)
bnj-贝叶斯网络java\examples\asia\asiaevidence.txt (149, 2003-03-18)
bnj-贝叶斯网络java\examples\asia\asiaProbs.txt (203, 2003-03-18)
bnj-贝叶斯网络java\examples\asia\asiatry.net (3440, 2003-03-18)
bnj-贝叶斯网络java\examples\asia\asia_train.xml (187228, 2003-03-18)
bnj-贝叶斯网络java\examples\asia\evidence.txt (158, 2003-03-18)
bnj-贝叶斯网络java\examples\asia\learnedbbn.xml (3500, 2003-03-18)
bnj-贝叶斯网络java\examples\genenet\yeast\cell-235-76.arff (106811, 2003-03-18)
bnj-贝叶斯网络java\examples\hailfinder\hailfinder.bif (55626, 2003-03-18)
bnj-贝叶斯网络java\examples\hailfinder\hailfinder.ergo (38446, 2003-03-18)
bnj-贝叶斯网络java\examples\hailfinder\hailfinder.net (51633, 2003-03-18)
bnj-贝叶斯网络java\examples\insurance\insurance.bif (21452, 2003-03-18)
bnj-贝叶斯网络java\examples\insurance\insurance.net (24106, 2003-03-18)
bnj-贝叶斯网络java\examples\link\Link.bif (246179, 2003-03-18)
bnj-贝叶斯网络java\examples\link\Link.hugin (263787, 2003-03-18)
bnj-贝叶斯网络java\examples\sprinkler\sprinkler.bif (518, 2003-03-18)
bnj-贝叶斯网络java\examples\sprinkler\sprinkler.bnt (869, 2003-03-18)
bnj-贝叶斯网络java\examples\university\university2.xml (4951, 2003-03-18)
bnj-贝叶斯网络java\examples\unsorted\comp_assym.net (12109, 2003-03-18)
bnj-贝叶斯网络java\examples\unsorted\dog-problem.xeb (343, 2003-03-18)
bnj-贝叶斯网络java\examples\unsorted\dog-problem.xml (2316, 2003-03-18)
... ...

Bayesian Network for Java 2.0 pre-alpha 20030318 ------------------------------------------------ This is Bayesian Network tools for Java developed by KDD Research (http://www.kddresearch.org). You can get the latest version at: http://bndev.sourceforge.net BNJ is licensed under GPL version 2 or any later version, at your option. Full text of the licence can be found in LICENSE.txt BNJ uses other libraries, including, but not limited to, OpenJGraph, Xerces, Log4J, and Ant. The license for these libraries can be found in their respective LICENSE files. We use Mersenne Twister fast implementation of Sean Luke's in package edu.ksu.cis.kdd.util, file MersenneTwisterFast.java. Note that this release is pre-alpha. A lot of features are still missing and/or contains severe bugs. We decided to pre-release this so that we can get as many feedbacks as possible and as early as possible. Requirements: ------------- * JDK version 1.4 or later, which can be downloaded at: http://java.sun.com * A computer and an operating system that can run Java. ISSUES ------ * For this pre-release, the GUI does not work properly yet. What it does is just loading the network and displaying it, nothing more. In the mean time, you should use the command line version instead. Follow the instruction below. * Currently, BNJ only accepts BIF, XML-BIF, XBN, and Hugin .NET format as network files. * For the data format, BNJ currently accept: - WEKA's .arff format - XML-BIF style data format - Comma separated data format - .DAT file. Preliminaries: -------------- First of all, you will need to set the environment variable BNJDIR to your BNJ installation directory. For example: For Windows XP/2000/NT4, if you installed BNJ to c:\bnj, then follow these steps: * Right click "My Computer" icon and choose "Properties". If there is no "My Computer" icon, click the Start button, choose "Settings", then "Control Panel", then double click "System" icon. * Click "Advanced" tab, then click "Environment Variables" button. * Notice there are two "New" buttons. Click the upper one. * Type in "BNJDIR" (without the quotes) in "Variable name". * Type in the BNJ installation directory in "Variable value". * Click OK, then OK again (in Environment Variable window), then OK again (in System Properties window). For Windows 9x or ME, consult the manual for details. For UNIX variant, you can set it up in your start up script (i.e. .profile or .cshrc). For Bash users, add in the following line in your .profile: export BNJDIR= For Csh or Tcsh users, add in the following line in your .cshrc file: setenv BNJDIR I. Instruction for Rebuilding from the Source --------------------------------------------- Firstly, make sure the BNJDIR environment variable is set. Then, invoke "build" script included with this package. For Windows user, use "build.bat" at all times, even if you use Cygwin. For UNIX or Linux users, use "build" script. Note that "javac" must be placed within the search path. II. Instruction for Running BNJ ------------------------------- Firstly, make sure the BNJDIR environment variable is set. Then, invoke "bnj" script included with this package. For Windows user, use "bnj.bat" at all times, even if you use Cygwin. For UNIX or Linux users, use "bnj" script. Note that "java" must be placed within the search path. If you invoke "bnj" script without parameters, it will take you straight to the GUI. NOTE: In this pre-release, the GUI doesn't work yet. III. Command Line BNJ --------------------- Invoking command line BNJ is similar to invoking the GUI -- using the "bnj" script with extra parameters. For help, type: bnj help Generally, the command line invocation format is as follows: bnj module_name parameters... For example: To invoke LS inference from a particular graph, type in: bnj ls -i:examples/asia/asia.xml If you want to invoke LS with an evidence file, type in: bnj ls -i:examples/asia/asia.xml -e:examples/asia/asia.xeb Below is the output of the command line help: ------------------------------------------------------------------------------------------------- Usage: edu.ksu.cis.bnj.BNJ [-r:randomseed] [-c:customconfigfile] [modulename] (moduleoptions)* -r = Setting the random seed (only if you want reproducability, otherwise it is set to system clock). -c = Load the specified configuration file name instead of the default. No module names will bring you to the GUI. Available modules are: Exact Inference: ls = Lauritzen-Spiegelhalter inference algorithm. elimbel = Variable elimination inference algorithm. pearl = Pearl's inference algorithm for singly connected networks Approximate Inference: ss = Simple sampling only for inference. logicsampling = Logic sampling for both inference and data generation. lw = Likelihood weighting for both inference and data generation. sis = Self importance sampling for both inference and data generation. pearlmcmc = Pearl MCMC method for both inference and data generation. chavez = Chavez MCMC for both inference and data generation. Structure Learning: k2 = K2 structure learning. gawk = Genetic Algorithm Wrapper for K2 [EXPERIMENTAL]. Data Generator: logicsampling = Logic sampling for both inference and data generation. lw = Likelihood weighting for both inference and data generation. sis = Self importance sampling for both inference and data generation. pearlmcmc = Pearl MCMC method for both inference and data generation. chavez = Chavez MCMC for both inference and data generation. Analysis: robustness = Robustness analysis for structure learning. Utilities: converter = Bayesian Network file format converter. dataconverter = Data file format converter. For individual module parameter helps, invoke the module without parameters. ------------------------------------------------------------------------------------------------- IV. Evidence File Format ------------------------ The new BNJ 2.0 features a new evidence file format. It is now XML-based and called XEB (XML-Based Evidence file for Bayesian Network). One example for Asia network is as follows: ------------------------------------------------------------------------------------------------- ]> ------------------------------------------------------------------------------------------------- As you may notice, the evidence file contains evidence of "Smoking = NonSmoker" and "Tuberculosis = Absent". The general format is in the tag as follows: V. To Get Further Help and/or Reporting Bugs -------------------------------------------- Please join BNJ user forum at: http://groups.yahoo.com/group/bndev Thank you.

近期下载者

相关文件


收藏者