root/bdm/estim/merger.cpp @ 283

#include "merger.h"
#include "arx.h"

namespace bdm{
vec merger::lognorm_merge ( mat &lW ) {
        int nu=lW.rows();
        vec mu = sum ( lW ) /nu; //mean of logs
        // vec lam = sum ( pow ( lW,2 ) )-nu*pow ( mu,2 ); ======= numerically unsafe!
        vec lam = sum ( pow ( lW-outer_product ( ones ( lW.rows() ),mu ),2 ) );
        double coef=0.0;
        vec sq2bl=sqrt ( 2*beta*lam ); //this term is everywhere
        switch ( nu ) {
                case 2:
                        coef= ( 1-0.5*sqrt ( ( 4.0*beta-3.0 ) /beta ) );
                        return  coef*sq2bl + mu ;
                        break;
                case 3://Ratio of Bessel
                        coef = sqrt ( ( 3*beta-2 ) /3*beta );
                        return log ( besselk ( 0,sq2bl*coef ) ) - log ( besselk ( 0,sq2bl ) ) +  mu;
                        break;
                case 4:
                        break;
                default: // Approximate conditional density
                        break;
        }
        return vec ( 0 );
}

void merger::merge ( const epdf* g0 ) {
//      it_file dbg ( "merger_debug.it" );

        it_assert_debug ( rv.equal ( g0->_rv() ),"Incompatible g0" );
        //Empirical density - samples
        eSmp.set_statistics ( ones ( Ns ), g0 );
        Array<vec> &Smp = eSmp._samples(); //aux
        vec &w = eSmp._w(); //aux

        mat Smp_ex =ones ( dim +1,Ns ); // Extended samples for the ARX model - the last row is ones
        for ( int i=0;i<Ns;i++ ) {      set_col_part ( Smp_ex,i,Smp ( i ) );}

//      dbg << Name ( "Smp_0" ) << Smp_ex;

        // Stuff for merging
        vec lw_src ( Ns );
        vec lw_mix ( Ns );
        vec lw ( Ns );
        mat lW=zeros ( n,Ns );
        vec vec0 ( 0 );

        // Initial component in the mixture model
        mat V0=1e-8*eye ( dim +1 );
        ARX A0; 
        A0.set_statistics(dim, V0, dim*dim +5.0 ); //initial guess of Mix: zero mean, large variance

        Mix.init ( &A0, Smp_ex, Nc );
        //Preserve initial mixture for repetitive estimation via flattening
        MixEF Mix_init ( Mix );

        // ============= MAIN LOOP ==================
        bool converged=false;
        int niter = 0;
        char str[100];

        epdf* Mpred=Mix.epredictor (  );
        vec Mix_pdf ( Ns );
        while ( !converged ) {
                //Re-estimate Mix
                //Re-Initialize Mixture model
                Mix.flatten ( &Mix_init );
                Mix.bayesB ( Smp_ex, w*Ns );
                delete Mpred;
                Mpred = Mix.epredictor ( ); // Allocation => must be deleted at the end!!

                // This will be active only later in iterations!!!
                if ( 1./sum_sqr ( w ) <0.5*Ns ) {
                        // Generate new samples
                        eSmp.set_samples ( Mpred );
                        for ( int i=0;i<Ns;i++ ) {
                                //////////// !!!!!!!!!!!!!
                                if ( Smp ( i ) ( 2 ) <0 ) {Smp ( i ) ( 2 ) = 0.01; }
                                set_col_part ( Smp_ex,i,Smp ( i ) );
                        }
                        if(0){cout<<"Resampling =" << 1./sum_sqr ( w ) << endl;
                        cout << sum ( Smp_ex,2 ) /Ns <<endl;
                        cout << Smp_ex*Smp_ex.T() /Ns << endl;}
                }
//              sprintf ( str,"Mpred_mean%d",niter );
//              dbg << Name ( str ) << Mpred->mean();


//              sprintf ( str,"Mpdf%d",niter );
//              for ( int i=0;i<Ns;i++ ) {Mix_pdf ( i ) = Mix.logpred ( Smp_ex.get_col ( i ) );}
//              dbg << Name ( str ) << Mix_pdf;

//              sprintf ( str,"Smp%d",niter );
//              dbg << Name ( str ) << Smp_ex;

                //Importace weighting
                for ( int i=0;i<n;i++ ) {
                        lw_src=0.0;
                        //======== Same RVs ===========
                        //Split according to dependency in rvs
                        if ( mpdfs ( i )->dimension() ==dim ) {
                                // no need for conditioning or marginalization
                                for ( int j=0;j<Ns; j++ ) { // Smp is Array<> => for cycle
                                        lw_src ( j ) =mpdfs ( i )->_epdf().evallog ( Smp ( j ) );
                                }
                        }
                        else {
                                // compute likelihood of marginal on the conditional variable
                                if ( mpdfs ( i )->dimensionc() >0 ) {
                                        // Make marginal on rvc_i
                                        epdf* tmp_marg = Mpred->marginal ( mpdfs ( i )->_rvc() );
                                        //compute vector of lw_src
                                        for ( int k=0;k<Ns;k++ ) {
                                                // Here val of tmp_marg = cond of mpdfs(i) ==> calling dls->get_cond
                                                lw_src ( k ) += tmp_marg->evallog ( dls ( i )->get_cond ( Smp ( k ) ) );
                                        }
                                        delete tmp_marg;

//                                      sprintf ( str,"marg%d",niter );
//                                      dbg << Name ( str ) << lw_src;

                                }
                                // Compute likelihood of the missing variable
                                if ( dim > ( mpdfs ( i )->dimension() + mpdfs ( i )->dimensionc() ) ) {
                                        ///////////////
                                        // There are variales unknown to mpdfs(i) : rvzs
                                        mpdf* tmp_cond = Mpred->condition ( rvzs ( i ) );
                                        // Compute likelihood
                                        vec lw_dbg=lw_src;
                                        for ( int k= 0; k<Ns; k++ ) {
                                                lw_src ( k ) += log (
                                                                    tmp_cond->evallogcond (
                                                                        zdls ( i )->pushdown ( Smp ( k ) ),
                                                                        zdls ( i )->get_cond ( Smp ( k ) ) ) );
                                                if ( !std::isfinite ( lw_src ( k ) ) ) {
                                                        lw_src ( k ) = -1e16; cout << "!";
                                                }
                                        }
                                        delete tmp_cond;
                                }
                                // Compute likelihood of the partial source
                                for ( int k= 0; k<Ns; k++ ) {
                                        mpdfs ( i )->condition ( dls ( i )->get_cond ( Smp ( k ) ) );
                                        lw_src ( k ) += mpdfs ( i )->_epdf().evallog ( dls ( i )->pushdown ( Smp ( k ) ) );
                                }

                        }
//                      it_assert_debug(std::isfinite(sum(lw_src)),"bad");
                        lW.set_row ( i, lw_src ); // do not divide by mix
                }
                //Importance of the mixture
                for ( int j=0;j<Ns;j++ ) {
                        lw_mix ( j ) =Mix.logpred ( Smp_ex.get_col ( j ) );
                }
//              sprintf ( str,"lW%d",niter );
//              dbg << Name ( str ) << lW;

                lw = lognorm_merge ( lW ); //merge

//              sprintf ( str,"w%d",niter );
//              dbg << Name ( str ) << w;
//              sprintf ( str,"lw_m%d",niter );
//              dbg << Name ( str ) << lw_mix;

                //Importance weighting
                lw -=  lw_mix; // hoping that it is not numerically sensitive...
                w = exp ( lw-max ( lw ) );
                //renormalize
                double sumw =sum ( w );
                if ( std::isfinite ( sumw ) ) {
                        w = w/sumw;
                }
                else {
                        it_file itf ( "merg_err.it" );
                        itf << Name ( "w" ) << w;
                }

//              sprintf ( str,"w_is_%d",niter );
//              dbg << Name ( str ) << w;

//              eSmp.resample(); // So that it can be used in bayes
//              for ( int i=0;i<Ns;i++ ) {      set_col_part ( Smp_ex,i,Smp ( i ) );}

//              sprintf ( str,"Smp_res%d",niter );
//              dbg << Name ( str ) << Smp;

                // ==== stopping rule ===
                niter++;
                converged = ( niter>20 );
        }
        delete Mpred;
        cout << endl;

}

}