Bayes net Management Tools
Ruby
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Failed to load latest commit information.
examples
lib
log
spec
.gitignore
.rspec
.ruby-version
CONTRIBUTING.md
Gemfile
LICENSE.txt
README.md
Rakefile
netica.gemspec

README.md

Netica

The Netica gem provides tools for interacting with Bayesian networks using JRuby and the Netica-J API, published by Norsys Software Corp..

Installation

Download the Netica-J API from Norsys and place the NeticaJ.jar file in your JRuby load path. Possible locations include...

/Library/Java/Extensions
/Network/Library/Java/Extensions
/System/Library/Java/Extensions
/usr/lib/java

Add this line to your application's Gemfile:

gem 'netica'

And then execute:

$ bundle install

Or install it yourself as:

$ gem install netica

Usage

Here's an example, based the Probabilistic Inference example in the Netica-J Manual.

Use Netica::Environment.engage to create a Netica Environ singleton object:

processor = Netica::Environment.engage
=> true

Create an ActiveNetwork, using a .dne file created in the Netica Application.

my_network = Netica::ActiveNetwork.new("some_identifiying_token", "./examples/ChestClinic.dne")
=> #<Netica::ActiveNetwork:0x58767e95 @network=#<Netica::BayesNetwork:0x4b709592 @current_network=#<Java::NorsysNetica::Net:0x75e82fc2>, @dne_file_path="./examples/ChestClinic.dne">, @token="some_identifiying_token">

View the nodes in the network.

my_network.network.nodes
=> #<Java::NorsysNetica::NodeList:1738447710 size:8 nodes:["NatureNode:VisitAsia", "NatureNode:Tuberculosis", "NatureNode:Smoking", "NatureNode:Cancer", "NatureNode:TbOrCa", "NatureNode:XRay", "NatureNode:Bronchitis", "NatureNode:Dyspnea"]>

Read the value of a Belief node.

tb_node = my_network.network.node("Tuberculosis")
=> #<Java::NorsysNetica::NatureNode:Tuberculosis value:{"present"=>0.010399998165667057, "absent"=>0.9896000027656555}>

tb_node.value("present")
=> 0.010399998165667057

Set the state of the XRay node to Abnormal.

xray_node = my_network.network.node("XRay")
=> #<Java::NorsysNetica::NatureNode:XRay value:{"abnormal"=>0.1102900430560112, "normal"=>0.8897099494934082}>

xray_node.value = "abnormal"
=> "abnormal"

Then, re-read the value of the Belief node.

p tb_node.value("present")
0.09241089224815369
=> 0.09241089224815369

Release Notes

Version 1.0.0.pre

  • Split ActiveNetwork's out-of-process storage and retrieval methods into a new subclass, StorableNetwork.
  • Add filename parameter to StorableNetwork#find, allowing a stored network's .dne file to be updated.
  • Built and tested with JRuby 1.7.9

Version 0.0.20

Delete a stored network by token with ActiveNetwork#destroy_by_token.

Version 0.0.19

Add load_from_storage parameter to ActiveNetwork#find so we can retrieve the state of a network in storage without instantiating it. Defaults to true.

Version 0.0.18

  • Return a hash from ActiveNetwork#destroy indicating the outcome of deletion in each storage location.
  • Make the java_library_path task visible in rake -vT

Version 0.0.17

  • We don't wait, we sleep when a network is locked.

Version 0.0.16

  • Add ActiveNetwork#destroy

Version 0.0.15

  • Lock in-use ActiveNetworks to prevent editing collisions

Version 0.0.14

  • User may specify optional container for Active Networks

Version 0.0.11

  • Move ActiveNetworks inside Environ

Version 0.0.10

  • Log Specifics of Node Changes

Version 0.0.9

  • Attempt to make Netica Environ a Singleton in JRuby 1.7

Version 0.0.8

  • Added Batch Processing as BayesNetwork#analyze, which takes a hash of node names and their values including an "id", and a list of output nodes, and returns the resulting values of the output nodes, along with the "id".

Version 0.0.7

  • Added Node Set identification and collection to BayesNetwork

Version 0.0.6

  • Changed BayesNetwork#getNode to BayesNetwork#node
  • Find a location for the NeticaJ.jar file using rake netica:java_library_path
  • Expand documentation

Legal

Netica and Norsys are registered trademarks of Norsys Software Corp.