eu.amidst.core.exponentialfamily.EF_SparseDirichlet.java Source code

Introduction

Here is the source code for eu.amidst.core.exponentialfamily.EF_SparseDirichlet.java

Source

/*
 * Licensed to the Apache Software Foundation (ASF) under one or more contributor license agreements.  See the NOTICE file distributed with this work for additional information regarding copyright ownership. The ASF licenses this file to You under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License.  You may obtain a copy of the License at
 *
 *       http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 *
 * See the License for the specific language governing permissions and limitations under the License.
 *
 */

package eu.amidst.core.exponentialfamily;

import eu.amidst.core.distribution.UnivariateDistribution;
import eu.amidst.core.utils.SparseVectorDefaultValue;
import eu.amidst.core.utils.Vector;
import eu.amidst.core.variables.Variable;
import org.apache.commons.math3.special.Gamma;

import java.util.ArrayList;
import java.util.Random;
import java.util.function.Function;

/**
 *
 * This class extends the abstract class {@link EF_UnivariateDistribution} and defines a Dirichlet distribution in Exponential Family (EF) canonical form.
 *
 * <p> For further details about how exponential family models are considered in this toolbox, take a look at the following paper:
 *  <i>Representation, Inference and Learning of Bayesian Networks as Conjugate Exponential Family Models. Technical Report.</i>
 * (<a href="http://amidst.github.io/toolbox/docs/ce-BNs.pdf">pdf</a>)
 * </p>
 */
public class EF_SparseDirichlet extends EF_UnivariateDistribution {

    /** Represents the number of parameter of this EF_Dirichlet distribution. */
    int nOfStates;

    /**
     * Creates a new uniform EF_Dirichlet distribution for a given {@link Variable} object.
     * @param var1 a {@link Variable} object.
     */
    public EF_SparseDirichlet(Variable var1) {
        if (!var1.isSparseDirichletParameter())
            throw new IllegalArgumentException("Non Dirichlet var");
        this.var = var1;
        this.nOfStates = var.getNumberOfStates();
        this.naturalParameters = this.createZeroNaturalParameters();
        this.momentParameters = this.createZeroMomentParameters();

        this.parents = new ArrayList();

        this.getSparseNaturalParameters().setDefaultValue(1.0);
        this.fixNumericalInstability();
        updateMomentFromNaturalParameters();
    }

    /**
     * Creates a new uniform EF_Dirichlet distribution for a given {@link Variable} object with a given scale.
     * @param var1 a {@link Variable} object with a Dirichlet distribution type
     * @param scale a positive double value defining the scale of the EF_Dirichlet.
     */
    public EF_SparseDirichlet(Variable var1, double scale) {
        if (!var1.isSparseDirichletParameter())
            throw new IllegalArgumentException("The variable is not a sparse Dirichlet parameter!");
        this.var = var1;
        this.nOfStates = var.getNumberOfStates();
        this.naturalParameters = this.createZeroNaturalParameters();
        this.momentParameters = this.createZeroMomentParameters();

        this.parents = new ArrayList();

        this.getSparseNaturalParameters().setDefaultValue(scale - 1.0);

        fixNumericalInstability();
        updateMomentFromNaturalParameters();
    }

    private SparseVectorDefaultValue getSparseNaturalParameters() {
        return ((SparseVectorDefaultValue) this.naturalParameters);
    }

    private SparseVectorDefaultValue getSparseMomentParameters() {
        return ((SparseVectorDefaultValue) this.momentParameters);
    }

    /**
     * {@inheritDoc}
     */
    @Override
    public double computeLogBaseMeasure(double val) {
        return 0;
    }

    /**
     * {@inheritDoc}
     */
    @Override
    public SufficientStatistics getSufficientStatistics(double val) {
        SufficientStatistics vec = this.createZeroSufficientStatistics();
        vec.set((int) val, Math.log(val));
        return vec;
    }

    /**
     * {@inheritDoc}
     */
    @Override
    public Vector getExpectedParameters() {

        double sum = this.naturalParameters.sum() + this.naturalParameters.size();

        SparseVectorDefaultValue vector = new SparseVectorDefaultValue(this.nOfStates, 0.0);
        vector.copy(this.naturalParameters);

        vector.sumConstant(1.0);
        vector.divideBy(sum);

        return vector;
    }

    /**
     * {@inheritDoc}
     */
    @Override
    public void fixNumericalInstability() {

    }

    /**
     * {@inheritDoc}
     */
    @Override
    public EF_UnivariateDistribution deepCopy(Variable var) {
        EF_SparseDirichlet copy = new EF_SparseDirichlet(var);
        copy.getNaturalParameters().copy(this.getNaturalParameters());
        copy.getMomentParameters().copy(this.getMomentParameters());
        return copy;
    }

    /**
     * {@inheritDoc}
     */
    @Override
    public EF_UnivariateDistribution randomInitialization(Random random) {
        for (int i = 0; i < this.nOfStates; i++) {
            this.getNaturalParameters().set(i, 5 * random.nextDouble() + 1 + 1e-5);
        }
        fixNumericalInstability();
        this.updateMomentFromNaturalParameters();

        return this;
    }

    /**
     * {@inheritDoc}
     */
    @Override
    public <E extends UnivariateDistribution> E toUnivariateDistribution() {
        throw new UnsupportedOperationException("Dirichlet is not included yet in the Distributions package.");
    }

    /**
     * {@inheritDoc}
     */
    @Override
    public void updateNaturalFromMomentParameters() {
        throw new UnsupportedOperationException(
                "No Implemented. EF_Dirichlet distribution should (right now) only be used for learning.");
    }

    /**
     * {@inheritDoc}
     */
    @Override
    public void updateMomentFromNaturalParameters() {

        double sumOfU_i = this.naturalParameters.sum();

        this.momentParameters = this.createZeroMomentParameters();

        this.momentParameters.copy(this.naturalParameters);

        this.getSparseMomentParameters().apply(new Function<Double, Double>() {
            @Override
            public Double apply(Double aDouble) {
                return Gamma.digamma(aDouble) - Gamma.digamma(sumOfU_i);
            }
        });

    }

    /**
     * {@inheritDoc}
     */
    @Override
    public int sizeOfSufficientStatistics() {
        return nOfStates;
    }

    /**
     * {@inheritDoc}
     */
    @Override
    public double computeLogNormalizer() {

        double sumOfU_i = this.naturalParameters.sum();

        double sumLogGammaOfU_i = this.getSparseNaturalParameters().sumApply(new Function<Double, Double>() {
            @Override
            public Double apply(Double aDouble) {
                return Gamma.logGamma(aDouble);
            }
        });

        return sumLogGammaOfU_i - Gamma.logGamma(sumOfU_i);

    }

    /**
     * {@inheritDoc}
     */
    @Override
    public Vector createZeroVector() {
        return new SparseVectorDefaultValue(nOfStates, 0.0);
    }

    /**
     * {@inheritDoc}
     */
    @Override
    public SufficientStatistics createInitSufficientStatistics() {
        return new SparseVectorDefaultValue(nOfStates,
                Gamma.digamma(1.0) - Gamma.digamma(this.sizeOfSufficientStatistics()));
    }
}