root/library/bdm/estim/mixtures.cpp @ 682

#include <vector>
#include "mixtures.h"

namespace bdm {


void MixEF::init ( BMEF* Com0, const mat &Data, int c ) {
        //prepare sizes
        Coms.set_size ( c );
        n = c;
        weights.set_parameters ( ones ( c ) ); //assume at least one observation in each comp.
        //est will be done at the end
        //
        int i;
        int ndat = Data.cols();
        //Estimate  Com0 from all data
        Coms ( 0 ) = Com0->_copy_();
//      Coms(0)->set_evalll(false);
        Coms ( 0 )->bayes_batch ( Data );
        // Flatten it to its original shape
        Coms ( 0 )->flatten ( Com0 );

        //Copy it to the rest
        for ( i = 1; i < n; i++ ) {
                //copy Com0 and create new rvs for them
                Coms ( i ) =  Coms ( 0 )->_copy_ ( );
        }
        //Pick some data for each component and update it
        for ( i = 0; i < n; i++ ) {
                //pick one datum
                int ind = floor ( ndat * UniRNG.sample() );
                Coms ( i )->bayes ( Data.get_col ( ind ), empty_vec );
                //flatten back to oringinal
                Coms ( i )->flatten ( Com0 );
        }

        //est already exists - must be deleted before build_est() can be used
        delete est;
        build_est();

}

void MixEF::bayes_batch ( const mat &data , const mat &cond, const vec &wData ) {
        int ndat = data.cols();
        int t, i, niter;
        bool converged = false;

        multiBM weights0 ( weights );

        Array<BMEF*> Coms0 ( n );
        for ( i = 0; i < n; i++ ) {
                Coms0 ( i ) = ( BMEF* ) Coms ( i )->_copy_();
        }

        niter = 0;
        mat W = ones ( n, ndat ) / n;
        mat Wlast = ones ( n, ndat ) / n;
        vec w ( n );
        vec ll ( n );
        // tmp for weights
        vec wtmp = zeros ( n );
        int maxi;
        double maxll;
        //Estim
        while ( !converged ) {
                // Copy components back to their initial values
                // All necessary information is now in w and Coms0.
                Wlast = W;
                //
                //#pragma omp parallel for
                for ( t = 0; t < ndat; t++ ) {
                        //#pragma omp parallel for
                        for ( i = 0; i < n; i++ ) {
                                ll ( i ) = Coms ( i )->logpred ( data.get_col ( t ) );
                                wtmp = 0.0;
                                wtmp ( i ) = 1.0;
                                ll ( i ) += weights.logpred ( wtmp );
                        }

                        maxll = max ( ll, maxi );
                        switch ( method ) {
                        case QB:
                                w = exp ( ll - maxll );
                                w /= sum ( w );
                                break;
                        case EM:
                                w = 0.0;
                                w ( maxi ) = 1.0;
                                break;
                        }

                        W.set_col ( t, w );
                }

                // copy initial statistics
                //#pragma omp parallel for
                for ( i = 0; i < n; i++ ) {
                        Coms ( i )-> set_statistics ( Coms0 ( i ) );
                }
                weights.set_statistics ( &weights0 );

                // Update statistics
                // !!!!    note  wData ==> this is extra weight of the data record
                // !!!!    For typical cases wData=1.
                for ( t = 0; t < ndat; t++ ) {
                        //#pragma omp parallel for
                        for ( i = 0; i < n; i++ ) {
                                Coms ( i )-> bayes_weighted ( data.get_col ( t ), empty_vec, W ( i, t ) * wData ( t ) );
                        }
                        weights.bayes ( W.get_col ( t ) * wData ( t ) );
                }

                niter++;
                //TODO better convergence rule.
                converged = ( niter > 10 );//( sumsum ( abs ( W-Wlast ) ) /n<0.1 );
        }

        //Clean Coms0
        for ( i = 0; i < n; i++ ) {
                delete Coms0 ( i );
        }
}

void MixEF::bayes ( const vec &data, const vec &cond=empty_vec ) {

};

void MixEF::bayes ( const mat &data, const vec &cond=empty_vec ) {
        this->bayes_batch ( data, cond, ones ( data.cols() ) );
};


double MixEF::logpred ( const vec &yt ) const {

        vec w = weights.posterior().mean();
        double exLL = 0.0;
        for ( int i = 0; i < n; i++ ) {
                exLL += w ( i ) * exp ( Coms ( i )->logpred ( yt ) );
        }
        return log ( exLL );
}

emix* MixEF::epredictor ( ) const {
        Array<shared_ptr<epdf> > pC ( n );
        for ( int i = 0; i < n; i++ ) {
                pC ( i ) = Coms ( i )->epredictor ( );
        }
        emix* tmp;
        tmp = new emix( );
        tmp->set_parameters ( weights.posterior().mean(), pC );
        return tmp;
}

void MixEF::flatten ( const BMEF* M2 ) {
        const MixEF* Mix2 = dynamic_cast<const MixEF*> ( M2 );
        bdm_assert_debug ( Mix2->n == n, "Different no of coms" );
        //Flatten each component
        for ( int i = 0; i < n; i++ ) {
                Coms ( i )->flatten ( Mix2->Coms ( i ) );
        }
        //Flatten weights = make them equal!!
        weights.set_statistics ( & ( Mix2->weights ) );
}
}