001 package aima.test.probabilitytest;
002
003 import java.util.Hashtable;
004
005 import junit.framework.TestCase;
006 import aima.probability.BayesNet;
007 import aima.probability.BayesNetNode;
008 import aima.probability.EnumerationAsk;
009 import aima.probability.Query;
010
011 /**
012 * @author Ravi Mohan
013 *
014 */
015
016 public class ProbabilitySamplingTest extends TestCase {
017
018 public void testPriorSample() {
019 BayesNet net = createWetGrassNetwork();
020 MockRandomizer r = new MockRandomizer(
021 new double[] { 0.5, 0.5, 0.5, 0.5 });
022 Hashtable table = net.getPriorSample(r);
023 assertEquals(4, table.keySet().size());
024 assertEquals(Boolean.TRUE, table.get("Cloudy"));
025 assertEquals(Boolean.FALSE, table.get("Sprinkler"));
026 assertEquals(Boolean.TRUE, table.get("Rain"));
027 assertEquals(Boolean.TRUE, table.get("WetGrass"));
028 }
029
030 public void testRejectionSample() {
031 BayesNet net = createWetGrassNetwork();
032 MockRandomizer r = new MockRandomizer(new double[] { 0.1 });
033 Hashtable<String, Boolean> evidence = new Hashtable<String, Boolean>();
034 evidence.put("Sprinkler", Boolean.TRUE);
035 double[] results = net.rejectionSample("Rain", evidence, 100, r);
036 assertEquals(1.0, results[0], 0.001);
037 assertEquals(0.0, results[1], 0.001);
038
039 }
040
041 public void testLikelihoodWeighting() {
042 MockRandomizer r = new MockRandomizer(
043 new double[] { 0.5, 0.5, 0.5, 0.5 });
044 BayesNet net = createWetGrassNetwork();
045 Hashtable<String, Boolean> evidence = new Hashtable<String, Boolean>();
046 evidence.put("Sprinkler", Boolean.TRUE);
047 double[] results = net.likelihoodWeighting("Rain", evidence, 1000, r);
048 // System.out.println(results[0] + " " + results[1]);
049 assertEquals(1.0, results[0], 0.001);
050 assertEquals(0.0, results[1], 0.001);
051 }
052
053 public void testMCMCask() {
054 BayesNet net = createWetGrassNetwork();
055 MockRandomizer r = new MockRandomizer(
056 new double[] { 0.5, 0.5, 0.5, 0.5 });
057
058 Hashtable<String, Boolean> evidence = new Hashtable<String, Boolean>();
059 evidence.put("Sprinkler", Boolean.TRUE);
060 double[] results = net.mcmcAsk("Rain", evidence, 1, r);
061 // System.out.println(results[0] + " " + results[1]);
062 assertEquals(0.333, results[0], 0.001);
063 assertEquals(0.666, results[1], 0.001);
064
065 }
066
067 public void testMCMCask2() {
068 BayesNet net = createWetGrassNetwork();
069 MockRandomizer r = new MockRandomizer(
070 new double[] { 0.5, 0.5, 0.5, 0.5 });
071
072 Hashtable<String, Boolean> evidence = new Hashtable<String, Boolean>();
073 evidence.put("Sprinkler", Boolean.TRUE);
074 double[] results = net.mcmcAsk("Rain", evidence, 1, r);
075 // System.out.println(results[0] + " " + results[1]);
076 assertEquals(0.333, results[0], 0.001);
077 assertEquals(0.666, results[1], 0.001);
078
079 }
080
081 public void testEnumerationAskinMCMC() {
082 BayesNet net = createWetGrassNetwork();
083 MockRandomizer r = new MockRandomizer(
084 new double[] { 0.5, 0.5, 0.5, 0.5 });
085 Hashtable<String, Boolean> evidence = new Hashtable<String, Boolean>();
086 evidence.put("Rain", Boolean.TRUE);
087 evidence.put("Sprinkler", Boolean.TRUE);
088 Query q = new Query("Cloudy", new String[] { "Sprinkler", "Rain" },
089 new boolean[] { true, true });
090 double[] results = EnumerationAsk.ask(q, net);
091 double[] results2 = net.mcmcAsk("Cloudy", evidence, 1000);
092 // System.out.println(results[0] + " " + results[1]);
093 // System.out.println(results2[0] + " " + results2[1]);
094
095 }
096
097 private BayesNet createWetGrassNetwork() {
098 BayesNetNode cloudy = new BayesNetNode("Cloudy");
099 BayesNetNode sprinkler = new BayesNetNode("Sprinkler");
100 BayesNetNode rain = new BayesNetNode("Rain");
101 BayesNetNode wetGrass = new BayesNetNode("WetGrass");
102
103 sprinkler.influencedBy(cloudy);
104 rain.influencedBy(cloudy);
105 wetGrass.influencedBy(rain, sprinkler);
106
107 cloudy.setProbability(true, 0.5);
108 sprinkler.setProbability(true, 0.10);
109 sprinkler.setProbability(false, 0.50);
110
111 rain.setProbability(true, 0.8);
112 rain.setProbability(false, 0.2);
113
114 wetGrass.setProbability(true, true, 0.99);
115 wetGrass.setProbability(true, false, 0.90);
116 wetGrass.setProbability(false, true, 0.90);
117 wetGrass.setProbability(false, false, 0.00);
118
119 BayesNet net = new BayesNet(cloudy);
120 return net;
121 }
122 }