nnfw/src/cluster.cpp

/********************************************************************************
 *  Neural Network Framework.                                                   *
 *  Copyright (C) 2005-2011 Gianluca Massera <emmegian@yahoo.it>                *
 *                                                                              *
 *  This program is free software; you can redistribute it and/or modify        *
 *  it under the terms of the GNU General Public License as published by        *
 *  the Free Software Foundation; either version 2 of the License, or           *
 *  (at your option) any later version.                                         *
 *                                                                              *
 *  This program is distributed in the hope that it will be useful,             *
 *  but WITHOUT ANY WARRANTY; without even the implied warranty of              *
 *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the               *
 *  GNU General Public License for more details.                                *
 *                                                                              *
 *  You should have received a copy of the GNU General Public License           *
 *  along with this program; if not, write to the Free Software                 *
 *  Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA  *
 ********************************************************************************/

#include "cluster.h"
#include "liboutputfunctions.h"
#include "configurationhelper.h"
#include <exception>

namespace farsa {

Cluster::Cluster( unsigned int numNeurons, QString name )
    : Updatable(name), inputdata(numNeurons,true), outputdata(numNeurons,true), stateDelegates() {
    //--- create the delegates for inputs and outputs vector
    setDelegateFor<Cluster, &Cluster::inputs>( "inputs" );
    setDelegateFor<Cluster, &Cluster::outputs>( "outputs" );
    this->numneurons = numNeurons;
    inputdataptr = &inputdata;
    inputdata.zeroing();
    outputdataptr = &outputdata;
    outputdata.zeroing();
    accOff = true;
    setNeedReset( false );
    // SigmoidFunction as Default
    setOutFunction(new SigmoidFunction( 1.0 ));
}

Cluster::Cluster( ConfigurationParameters& params, QString prefix ) :
    Updatable(params, prefix), stateDelegates()
{
    //--- create the delegates for inputs and outputs vector
    setDelegateFor<Cluster, &Cluster::inputs>( "inputs" );
    setDelegateFor<Cluster, &Cluster::outputs>( "outputs" );
    numneurons = ConfigurationHelper::getInt( params, prefix + "numNeurons", 1 );
    // Creating inputdata and outputdata
    inputdata = DoubleVector(numneurons, true );
    inputdataptr = &inputdata;
    inputdata.zeroing();
    outputdata = DoubleVector(numneurons, true );
    outputdataptr = &outputdata;
    outputdata.zeroing();

    accOff = !ConfigurationHelper::getBool( params, prefix + "accumulate" );

    QString vectorSizeErrorTmpl( "The number of elements of the %1 vector in configuration file (%1) is different from the number of neurons (%2)");
    // inputs is a vector, that is a list of space-separated values
    QVector<double> vect = ConfigurationHelper::getVector( params, prefix + "inputs" );
#ifdef FARSA_DEBUG
    if ( !vect.isEmpty() && vect.size() != (int)numneurons ) {
        qWarning() << vectorSizeErrorTmpl.arg( "inputs" ).arg( vect.size() ).arg( numneurons );
    }
#endif
    inputdataptr->copyValues( vect );

    // outputs is a vector, that is a list of space-separated values
    vect = ConfigurationHelper::getVector( params, prefix + "outputs" );
#ifdef FARSA_DEBUG
    if ( !vect.isEmpty() && vect.size() != (int)numneurons ) {
        qWarning() << vectorSizeErrorTmpl.arg( "outputs" ).arg( vect.size() ).arg( numneurons );
    }
#endif
    outputdataptr->copyValues( vect );

    // The group for the outputFunction is "prefix/OutFunction"
    try {
        setOutFunction(params.getObjectFromGroup<OutputFunction>( prefix + "OutFunction", true ));
    } catch( std::exception& /*e*/) {
        // SigmoidFunction as Default
        setOutFunction(new SigmoidFunction( 1.0 ));
    }
    setNeedReset( false );
}

Cluster::~Cluster() {
    // No need to delete anything else, we use auto_ptr
}

void Cluster::setOutFunction( OutputFunction *up ) {
    updater.reset(up);
    updater->setCluster( this );
}

void Cluster::setInput( unsigned int neuron, double value ) {
    (*inputdataptr)[neuron] = value;
}

void Cluster::setInputs( const DoubleVector& inputs ) {
    inputdataptr->copyValues( inputs );
}

void Cluster::setAllInputs( double value ) {
    inputdataptr->setAll( value );
    setNeedReset( false );
}

void Cluster::resetInputs() {
    inputdataptr->zeroing();
    setNeedReset( false );
}

double Cluster::getInput( unsigned int neuron ) const {
    return (*inputdataptr)[neuron];
}

void Cluster::setOutput( unsigned int neuron, double value ) {
    (*outputdataptr)[neuron] = value;
}

void Cluster::setOutputs( const DoubleVector& outputs ) {
    outputdataptr->copyValues( outputs );
}

double Cluster::getOutput( unsigned int neuron ) const {
    return (*outputdataptr)[neuron];
}

void Cluster::save(ConfigurationParameters& params, QString prefix)
{
    Updatable::save( params, prefix );
    params.startObjectParameters(prefix, "Cluster", this);
    params.createParameter(prefix, "numNeurons", QString::number(numneurons));
    params.createParameter(prefix, "accumulate", (isAccumulate() ? "True" : "False"));
    // First creating a string list, then transforming to a single string
    QStringList list;
    for (unsigned int i = 0; i < inputdataptr->size(); i++) {
        list.push_back(QString::number((*inputdataptr)[i]));
    }
    params.createParameter(prefix, "inputs", list.join(" "));
    // Doing the same with outputdata
    list.clear();
    for (unsigned int i = 0; i < outputdataptr->size(); i++) {
        list.push_back(QString::number((*outputdataptr)[i]));
    }
    params.createParameter(prefix, "outputs", list.join(" "));
    // and finally the outfunction will be saved in the group "prefix/OutFunction"
    updater->save( params, params.createSubGroup(prefix, "OutFunction") );
}

void Cluster::describe( QString type ) {
    Updatable::describe( type );
    Descriptor d = addTypeDescription( type, "A Cluster of neurons" );
    d.describeInt( "numNeurons" ).limits( 1, INT_MAX ).props( IsMandatory ).help( "The number of neurons contained by the Cluster" );
    d.describeBool( "accumulate" ).def( false ).help( "If true new inputs will be added to the previous values" );
    d.describeReal( "inputs" ).props( IsList ).help( "The input values of the neurons" );
    d.describeReal( "outputs" ).props( IsList ).help( "The output values of the neurons" );
    d.describeObject( "OutFunction" ).type( "OutputFunction" ).help( "The output function used to calculate the output values" );
}

}