/** {@inheritDoc} */
@Override
public BayesianNetwork getLearntBayesianNetwork() {
// Normalize the sufficient statistics
SufficientStatistics normalizedSS = efBayesianNetwork.createZeroSufficientStatistics();
normalizedSS.copy(sumSS);
normalizedSS.divideBy(dataInstanceCount.get());
efBayesianNetwork.setMomentParameters(normalizedSS);
return efBayesianNetwork.toBayesianNetwork(dag);
}
/** {@inheritDoc} */
@Override
public void runLearning() {
this.initLearning();
efBayesianNetwork = new EF_BayesianNetwork(dag);
Stream<DataInstance> stream = null;
if (parallelMode) {
stream = dataStream.parallelStream(batchSize);
} else {
stream = dataStream.stream();
}
dataInstanceCount = new AtomicDouble(0);
sumSS =
stream
.peek(
w -> {
dataInstanceCount.getAndAdd(1.0);
})
.map(efBayesianNetwork::getSufficientStatistics)
.reduce(
efBayesianNetwork.createZeroSufficientStatistics(),
SufficientStatistics::sumVector);
}
public boolean equal_efBN(EF_BayesianNetwork ef_bayesianNetwork, double threshold) {
for (EF_ConditionalDistribution this_dist : this.getDistributionList()) {
EF_ConditionalDistribution ef_dist =
ef_bayesianNetwork.getDistribution(this_dist.getVariable());
if (!this_dist.getClass().getName().equals(ef_dist.getClass().getName())) return false;
List<Variable> this_Vars = this_dist.getConditioningVariables();
List<Variable> ef_Vars = ef_dist.getConditioningVariables();
if (this_Vars.size() != ef_Vars.size()) return false;
for (Variable var : this_Vars) {
if (!ef_Vars.contains(var)) return false;
}
}
return this.getNaturalParameters()
.equalsVector(ef_bayesianNetwork.getNaturalParameters(), threshold);
}
/** {@inheritDoc} */
@Override
public void initLearning() {
dataInstanceCount = new AtomicDouble(0);
efBayesianNetwork = new EF_BayesianNetwork(dag);
sumSS = efBayesianNetwork.createZeroSufficientStatistics();
}