Examples with ParameterEstimationForHOGModel com.sri.ai.praise.learning.symbolicparameterestimation.ParameterEstimationForHOGModel used on opensource projects

Example 1 with ParameterEstimationForHOGModel

use of com.sri.ai.praise.learning.symbolicparameterestimation.ParameterEstimationForHOGModel in project aic-praise by aic-sri-international.

the class ParameterEstimationForHOGModelTest method testHOGMBased.

// TODO: not matching original expected result put in by Sarah
// @Test
public void testHOGMBased() {
    String modelString = "random terrorAttacks : 0..20;\n" + "random newJobs : 0..100000;\n" + "random dow: 11000..18000;\n" + "random economyIsPoor : Boolean;\n" + "random economyIsGreat : Boolean;\n" + "random attackPerception: Boolean;\n" + "random likeIncumbent  : 0..100000000;\n" + "random likeChallenger : 0..100000000;\n" + "constant Alpha: Real;\n" + "economyIsPoor <=> dow < 13000 and newJobs < 30000;\n" + "economyIsGreat <=> dow > 16000 and newJobs > 70000;\n" + "attackPerception <=> terrorAttacks > 4;\n" + "if economyIsGreat\n" + "then if likeIncumbent > 70000000 then Alpha/30000000 else (1-Alpha)/(70000000 + 1)\n" + "else if economyIsPoor\n" + "then if likeIncumbent < 40000000 then 0.8/40000000 else 0.2/(60000000 + 1)\n" + "else if attackPerception\n" + "then if likeIncumbent < 60000000 then 0.9/60000000 else 0.1/(40000000 + 1);\n";
    List<HOGMProblemError> modelErrors = new ArrayList<>();
    List<Pair<Expression, Expression>> pairsQueryEvidence = new LinkedList<Pair<Expression, Expression>>();
    Pair<Expression, Expression> pair = new Pair<Expression, Expression>(parse("likeIncumbent > likeChallenger"), parse("null"));
    pairsQueryEvidence.add(pair);
    ParameterEstimationForHOGModel parameterEstimationForHOGModel = new ParameterEstimationForHOGModel(modelString, pairsQueryEvidence, modelErrors);
    HashMap<Expression, Double> expected = new HashMap<Expression, Double>();
    expected.put(parse("Alpha"), 1.0);
    HashMap<Expression, Double> mapResult = runTestHOGModelBased(expected, parameterEstimationForHOGModel, new double[] { 0 });
    HOGModel test = parameterEstimationForHOGModel.buildOptimizedHOGModel(mapResult);
    ExpressionBasedModel newModel = parseHOGModelToExpressionBasedModel(test);
    System.out.println(newModel.toString());
    assertEquals(expected, mapResult);
}

Example 2 with ParameterEstimationForHOGModel

use of com.sri.ai.praise.learning.symbolicparameterestimation.ParameterEstimationForHOGModel in project aic-praise by aic-sri-international.

the class ParameterEstimationForHOGModelTest method testBuildOptimizedHOGModel.

@Test
public void testBuildOptimizedHOGModel() {
    String modelString = "random earthquake: Boolean;\n" + "random burglary: Boolean;\n" + "random alarm: Boolean;\n" + "constant Alpha: Real;\n" + "constant Beta: Real;\n" + "earthquake Alpha;\n" + "burglary Beta;\n" + "if earthquake\n" + "then if burglary\n" + "then alarm 0.95\n" + "else alarm 0.6\n" + "else if burglary\n" + "then alarm 0.9\n" + "else alarm 0.01;\n";
    List<HOGMProblemError> modelErrors = new ArrayList<>();
    List<Pair<Expression, Expression>> pairsQueryEvidence = new LinkedList<Pair<Expression, Expression>>();
    Pair<Expression, Expression> pair = new Pair<Expression, Expression>(parse("earthquake"), parse("null"));
    pairsQueryEvidence.add(pair);
    ParameterEstimationForHOGModel parameterEstimationForHOGModel = new ParameterEstimationForHOGModel(modelString, pairsQueryEvidence, modelErrors);
    HashMap<Expression, Double> expected = new HashMap<Expression, Double>();
    expected.put(parse("Alpha"), 1.0);
    HashMap<Expression, Double> mapResult = runTestHOGModelBased(expected, parameterEstimationForHOGModel, new double[] { 0 });
    String newStringModel = parameterEstimationForHOGModel.buildOptimizedStringModel(mapResult);
    // ExpressionBasedModel newExpressionBasedModel = parameterEstimationForHOGModel.parseHOGModelToExpressionBasedModel(newStringModel);
    // System.out.println(newExpressionBasedModel.toString());
    System.out.println(newStringModel);
    pairsQueryEvidence.clear();
    pair = new Pair<Expression, Expression>(parse("burglary"), parse("null"));
    pairsQueryEvidence.add(pair);
    ParameterEstimationForHOGModel parameterEstimationForHOGModel2 = new ParameterEstimationForHOGModel(newStringModel, pairsQueryEvidence, modelErrors);
    expected.remove(parse("Alpha"));
    expected.put(parse("Beta"), 1.0);
    HashMap<Expression, Double> mapResult2 = runTestHOGModelBased(expected, parameterEstimationForHOGModel2, new double[] { 0 });
    System.out.println(mapResult);
    String newStringModel2 = parameterEstimationForHOGModel2.buildOptimizedStringModel(mapResult2);
    // ExpressionBasedModel newModel = parameterEstimationForHOGModel.parseHOGModelToExpressionBasedModel(test);
    System.out.println(newStringModel2.toString());
}