gov.sandia.cognition.learning.function.vector.DifferentiableGeneralizedLinearModel Maven / Gradle / Ivy

Go to download
Show more of this group Show more artifacts with this name
Show all versions of cognitive-foundry Show documentation
A single jar with all the Cognitive Foundry components.
There is a newer version: 4.0.1
/*
 * File:                DifferentiableGeneralizedLinearModel.java
 * Authors:             Kevin R. Dixon
 * Company:             Sandia National Laboratories
 * Project:             Cognitive Foundry
 *
 * Copyright February 28, 2007, Sandia Corporation.  Under the terms of Contract
 * DE-AC04-94AL85000, there is a non-exclusive license for use of this work by
 * or on behalf of the U.S. Government. Export of this program may require a
 * license from the United States Government. See CopyrightHistory.txt for
 * complete details.
 *
 */

package gov.sandia.cognition.learning.function.vector;

import gov.sandia.cognition.learning.algorithm.gradient.GradientDescendable;
import gov.sandia.cognition.learning.function.scalar.IdentityScalarFunction;
import gov.sandia.cognition.math.DifferentiableUnivariateScalarFunction;
import gov.sandia.cognition.math.matrix.DifferentiableVectorFunction;
import gov.sandia.cognition.math.matrix.Matrix;
import gov.sandia.cognition.math.matrix.Vector;

/**
 * A GradientDescenable version of a GeneralizedLinearModel, in
 * other words, a GeneralizedLinearModel where the squashing
 * function is differentiable
 *
 * @author Kevin R. Dixon
 * @since  1.0
 *
 */
public class DifferentiableGeneralizedLinearModel
    extends GeneralizedLinearModel
    implements GradientDescendable, DifferentiableVectorFunction
{

    /**
     * Default Constructor. Creates a 1x1 model with a identity function for
     * the output.
     */
    public DifferentiableGeneralizedLinearModel()
    {
        this( 1, 1, new IdentityScalarFunction() );
    }

    /**
     * Creates a new instance of GeneralizedLinearModel
     * 
     * @param numInputs
     * Number of inputs of the function (number of matrix columns)
     * @param numOutputs
     * Number of outputs of the function (number of matrix rows)
     * @param scalarFunction 
     * Function to apply to each output
     */
    public DifferentiableGeneralizedLinearModel(
        int numInputs,
        int numOutputs,
        DifferentiableUnivariateScalarFunction scalarFunction )
    {
        this( new MultivariateDiscriminant( numInputs, numOutputs ),
            new ElementWiseDifferentiableVectorFunction( scalarFunction ) );
    }
    
    
    /**
     * Creates a new instance of DifferentiableGeneralizedLinearModel
     * @param matrixMultiply 
     * GradientDescendable that multiplies an input by the internal matrix
     * @param squashingFunction 
     * VectorFunction that is applied to the output of the matrix multiply
     */
    public DifferentiableGeneralizedLinearModel(
        MultivariateDiscriminant matrixMultiply,
        DifferentiableVectorFunction squashingFunction )
    {
        super( matrixMultiply, squashingFunction );
    }

    /**
     * Creates a new instance of DifferentiableGeneralizedLinearModel
     * @param matrixMultiply 
     * GradientDescendable that multiplies an input by the internal matrix
     * @param scalarSquashingFunction 
     * scalar function that is applied to the output of the matrix multiply
     */
    public DifferentiableGeneralizedLinearModel(
        MultivariateDiscriminant matrixMultiply,
        DifferentiableUnivariateScalarFunction scalarSquashingFunction )
    {
        this( matrixMultiply, new ElementWiseDifferentiableVectorFunction(
            scalarSquashingFunction ) );
    }

    /**
     * Creates a new instance of DifferentiableGeneralizedLinearModel
     * 
     * @param other DifferentiableGeneralizedLinearModel to copy
     */
    public DifferentiableGeneralizedLinearModel(
        DifferentiableGeneralizedLinearModel other )
    {
        super( other );
    }

    @Override
    public DifferentiableVectorFunction getSquashingFunction()
    {
        return (DifferentiableVectorFunction) super.getSquashingFunction();
    }

    public Matrix computeParameterGradient(
        Vector input )
    {
        Matrix gradient =
            this.getDiscriminant().computeParameterGradient( input );

        Vector y = this.getDiscriminant().evaluate( input );
        Matrix derivative = this.getSquashingFunction().differentiate( y );

        return derivative.times( gradient );
    }

    @Override
    public DifferentiableGeneralizedLinearModel clone()
    {
        return (DifferentiableGeneralizedLinearModel) super.clone();
    }

    public Matrix differentiate(
        Vector input )
    {
        Matrix dudx = this.getDiscriminant().differentiate( input );

        Vector u = this.getDiscriminant().evaluate( input );
        Matrix dydu = this.getSquashingFunction().differentiate( u );

        return dydu.times( dudx );

    }

}