Construction & inference in C#

// --------------------------------------------------------------------------------------------------------------------
// <copyright file="NetworkExample.cs" company="Bayes Server">
// --------------------------------------------------------------------------------------------------------------------

namespace BayesServer.HelpSamples
{
using System;

using BayesServer.Inference.RelevanceTree;

public static class NetworkExample
{
public static void Main()
{
// In this example we programatically create a simple Bayesian network.
// Note that you can automatically define nodes from data using
// classes in BayesServer.Data.Discovery,
// and you can automatically learn the parameters using classes in
// BayesServer.Learning.Parameters,
// however here we build a Bayesian network from scratch.

var network = new Network("Demo");

// add the nodes (variables)

var aTrue = new State("True");
var aFalse = new State("False");
var a = new Node("A", aTrue, aFalse);

var bTrue = new State("True");
var bFalse = new State("False");
var b = new Node("B", bTrue, bFalse);

var cTrue = new State("True");
var cFalse = new State("False");
var c = new Node("C", cTrue, cFalse);

var dTrue = new State("True");
var dFalse = new State("False");
var d = new Node("D", dTrue, dFalse);

// at this point we have fully specified the structural (graphical) specification of the Bayesian Network.

// We must define the necessary probability distributions for each node.

// Each node in a Bayesian Network requires a probability distribution conditioned on it's parents.

// NewDistribution() can be called on a Node to create the appropriate probability distribution for a node
// or it can be created manually.

// The interface IDistribution has been designed to represent both discrete and continuous variables,

// As we are currently dealing with discrete distributions, we will use the
// Table class.

// To access the discrete part of a distribution, we use IDistribution.Table.

// The Table class is used to define distributions over a number of discrete variables.

var tableA = a.NewDistribution().Table;     // access the table property of the Distribution

// IMPORTANT
// Note that calling Node.NewDistribution() does NOT assign the distribution to the node.
// A distribution cannot be assigned to a node until it is correctly specified.
// If a distribution becomes invalid  (e.g. a parent node is added), it is automatically set to null.

tableA[aTrue] = 0.1;
tableA[aFalse] = 0.9;

// now tableA is correctly specified we can assign it to Node A;
a.Distribution = tableA;

// node B has node A as a parent, therefore its distribution will be P(B|A)

var tableB = b.NewDistribution().Table;
tableB[aTrue, bTrue] = 0.2;
tableB[aTrue, bFalse] = 0.8;
tableB[aFalse, bTrue] = 0.15;
tableB[aFalse, bFalse] = 0.85;
b.Distribution = tableB;

// specify P(C|A)
var tableC = c.NewDistribution().Table;
tableC[aTrue, cTrue] = 0.3;
tableC[aTrue, cFalse] = 0.7;
tableC[aFalse, cTrue] = 0.4;
tableC[aFalse, cFalse] = 0.6;
c.Distribution = tableC;

// specify P(D|B,C)
var tableD = d.NewDistribution().Table;

// we could specify the values individually as above, or we can use a TableIterator as follows
var iteratorD = new TableIterator(tableD, new Node[] { b, c, d });
iteratorD.CopyFrom(new double[] { 0.4, 0.6, 0.55, 0.45, 0.32, 0.68, 0.01, 0.99 });
d.Distribution = tableD;

// The network is now fully specified

// If required the network can be saved...

if (false)   // change this to true to save the network
{
// network.Save("fileName.bayes");  // replace 'fileName.bayes' with your own path and uncomment start of line
}

// Now we will calculate P(A|D=True), i.e. the probability of A given the evidence that D is true

// use the factory design pattern to create the necessary inference related objects
var factory = new RelevanceTreeInferenceFactory();
var inference = factory.CreateInferenceEngine(network);
var queryOptions = factory.CreateQueryOptions();
var queryOutput = factory.CreateQueryOutput();

// we could have created these objects explicitly instead, but as the number of algorithms grows
// this makes it easier to switch between them

inference.Evidence.SetState(dTrue);  // set D = True

var queryA = new Table(a);
inference.Query(queryOptions, queryOutput); // note that this can raise an exception (see help for details)

Console.WriteLine("P(A|D=True) = {" + queryA[aTrue] + "," + queryA[aFalse] + "}.");

// Expected output ...
// P(A|D=True) = {0.0980748663101604,0.90192513368984}

// to perform another query we reuse all the objects

// now lets calculate P(A|D=True, C=True)
inference.Evidence.SetState(cTrue);

// we will also return the log-likelihood of the case
queryOptions.LogLikelihood = true; // only request the log-likelihood if you really need it, as extra computation is involved

inference.Query(queryOptions, queryOutput);
Console.WriteLine(string.Format("P(A|D=True, C=True) = [{0},{1}], log-likelihood = {2}.", queryA[aTrue], queryA[aFalse], queryOutput.LogLikelihood.Value));

// Expected output ...
// P(A|D=True, C=True) = {0.0777777777777778,0.922222222222222}, log-likelihood = -2.04330249506396.

// Note that we can also calculate joint queries such as P(A,B|D=True,C=True)

}
}
}