Examples with InferenceForFactorGraphAndEvidence com.sri.ai.praise.sgsolver.solver.InferenceForFactorGraphAndEvidence used on opensource projects

Example 6 with InferenceForFactorGraphAndEvidence

use of com.sri.ai.praise.sgsolver.solver.InferenceForFactorGraphAndEvidence in project aic-praise by aic-sri-international.

the class InferenceForFactorGraphAndEvidenceTest method testBurglary.

@Test
public void testBurglary() {
    // The definitions of types
    mapFromCategoricalTypeNameToSizeString = Util.map("Boolean", "2");
    // The definitions of variables
    mapFromRandomVariableNameToTypeName = Util.map("burglary", "Boolean", "alarm", "Boolean", "call", "Boolean");
    // The definitions of non-uniquely named constants
    mapFromNonUniquelyNamedConstantNameToTypeName = Util.map();
    // The definitions of non-uniquely named constants
    mapFromUniquelyNamedConstantNameToTypeName = Util.map();
    isBayesianNetwork = false;
    factors = Times.getMultiplicands(parse("" + "(if alarm then if call then 0.7 else 0.3 else if call then 0 else 1)*" + "(if burglary then if alarm then 0.9 else 0.1 else if alarm then 0.01 else 0.99)*" + "(if burglary then 0.1 else 0.9)"));
    InferenceForFactorGraphAndEvidence inferencer;
    inferencer = new InferenceForFactorGraphAndEvidence(new ExpressionFactorsAndTypes(factors, mapFromRandomVariableNameToTypeName, mapFromNonUniquelyNamedConstantNameToTypeName, mapFromUniquelyNamedConstantNameToTypeName, mapFromCategoricalTypeNameToSizeString, list()), isBayesianNetwork, evidence, false, null);
    Expression result = inferencer.sum(list(parse("alarm")), Times.make(factors));
    System.out.println(result);
}

Example 7 with InferenceForFactorGraphAndEvidence

use of com.sri.ai.praise.sgsolver.solver.InferenceForFactorGraphAndEvidence in project aic-praise by aic-sri-international.

the class InferenceForFactorGraphAndEvidenceTest method runTestWithFactorizationOption.

/**
	 * @param useFactorization
	 * @param queryExpression
	 * @param evidence
	 * @param expected
	 * @param isBayesianNetwork
	 * @param factorGraph
	 * @param mapFromRandomVariableNameToTypeName
	 * @param mapFromNonUniquelyNamedConstantNameToTypeName
	 * @param mapFromUniquelyNamedConstantNameToTypeName
	 * @param mapFromCategoricalTypeNameToSizeString
	 */
private void runTestWithFactorizationOption(boolean useFactorization, Expression queryExpression, Expression evidence, Expression expected, boolean isBayesianNetwork, List<Expression> factors, Map<String, String> mapFromRandomVariableNameToTypeName, Map<String, String> mapFromNonUniquelyNamedConstantNameToTypeName, Map<String, String> mapFromUniquelyNamedConstantNameToTypeName, Map<String, String> mapFromCategoricalTypeNameToSizeString, Collection<Type> additionalTypes) {
    InferenceForFactorGraphAndEvidence inferencer;
    Expression marginal;
    inferencer = new InferenceForFactorGraphAndEvidence(new ExpressionFactorsAndTypes(factors, mapFromRandomVariableNameToTypeName, mapFromNonUniquelyNamedConstantNameToTypeName, mapFromUniquelyNamedConstantNameToTypeName, mapFromCategoricalTypeNameToSizeString, additionalTypes), isBayesianNetwork, evidence, useFactorization, null);
    marginal = inferencer.solve(queryExpression);
    TrueContext context = new TrueContext();
    marginal = Expressions.roundToAGivenPrecision(marginal, 9, context);
    expected = Expressions.roundToAGivenPrecision(expected, 9, context);
    if (expected.equals(marginal)) {
    // Ok!
    } else // check if they are not identical, but equivalent expressions
    if (inferencer.evaluate(apply(MINUS, expected, marginal)).equals(ZERO)) {
    // first attempt was to compare with equality, but this requires a more complete test of equality theory literals to exclude such a complex equality from being considered a literal, which is much more expensive
    // Ok!
    } else {
        throw new AssertionError("expected:<" + expected + "> but was:<" + marginal + ">, which is not even equivalent.");
    }
// Not working yet, need to debug		
//		Expression negationMarginal;
//		negationMarginal = inferencer.solve(Not.make(queryExpression));
//		negationMarginal = Expressions.roundToAGivenPrecision(negationMarginal, 9, context);
//		expected = inferencer.evaluate(parse(negationMarginal + " = 1 - " + marginal));
//		assertEquals(expected, TRUE);
}