root/library/bdm/stat/exp_family.cpp @ 1086

#include <math.h>

#include <itpp/base/bessel.h>
#include "exp_family.h"

namespace bdm {

Uniform_RNG UniRNG;
Normal_RNG NorRNG;
Gamma_RNG GamRNG;

using std::cout;

///////////

void BMEF::bayes ( const vec &yt, const vec &cond ) {
    this->bayes_weighted ( yt, cond, 1.0 );
};

void egiw::set_parameters ( int dimx0, ldmat V0, double nu0 ) {
    dimx = dimx0;
    nPsi = V0.rows()-dimx;
    V = V0;
    if ( nu0 < 0 ) {
        nu = 0.1 + nPsi + 2 * dimx + 2; // +2 assures finite expected value of R
        // terms before that are sufficient for finite normalization
    } else {
        nu = nu0;
    }
}

vec egiw::sample() const {
    mat M;
    chmat R;
    sample_mat ( M, R );

    return concat ( cvectorize ( M ), cvectorize ( R.to_mat() ) );
}

mat egiw::sample_mat ( int n ) const {
    // TODO - correct approach - convert to product of norm * Wishart
    mat M;
    ldmat Vz;
    ldmat Lam;
    factorize ( M, Vz, Lam );

    chmat ChLam ( Lam.to_mat() );
    chmat iChLam;
    ChLam.inv ( iChLam );

    eWishartCh Omega; //inverse Wishart, result is R,
    Omega.set_parameters ( iChLam, nu - 2*nPsi - dimx ); // 2*nPsi is there to match numercial simulations - check if analytically correct
    Omega.validate();

    mat OmChi;
    mat Z ( M.rows(), M.cols() );

    mat Mi;
    mat RChiT;
    mat tmp ( dimension(), n );
    M=M.T();// ugly hack == decide what to do with M.
    for ( int i = 0; i < n; i++ ) {
        OmChi = Omega.sample_mat();
        RChiT = inv ( OmChi );
        Z = randn ( M.rows(), M.cols() );
        Mi = M + RChiT * Z * inv ( Vz._L().T() * diag ( sqrt ( Vz._D() ) ) );

        tmp.set_col ( i, concat ( cvectorize ( Mi ), cvectorize ( RChiT*RChiT.T() ) ) );
    }
    return tmp;
}

void egiw::sample_mat ( mat &Mi, chmat &Ri ) const {

    // TODO - correct approach - convert to product of norm * Wishart
    mat M;
    ldmat Vz;
    ldmat Lam;
    factorize ( M, Vz, Lam );

    chmat Ch;
    Ch.setCh ( Lam._L() *diag ( sqrt ( Lam._D() ) ) );
    chmat iCh;
    Ch.inv ( iCh );

    eWishartCh Omega; //inverse Wishart, result is R,
    Omega.set_parameters ( iCh, nu - 2*nPsi - dimx ); // 2*nPsi is there to match numercial simulations - check if analytically correct
    Omega.validate();

    chmat Omi;
    Omi.setCh ( Omega.sample_mat() );

    if (M._datasize()>0) {
        mat Z = randn ( M.rows(), M.cols() );
        Mi = M + Omi._Ch() * Z * inv ( Vz._L() * diag ( sqrt ( Vz._D() ) ) );
    }
    Omi.inv ( Ri );
}

double egiw::evallog_nn ( const vec &val ) const {
    bdm_assert_debug(val.length()==dimx*(nPsi+dimx),"Incorrect cond in egiw::evallog_nn" );

    int vend = val.length() - 1;

    if ( dimx == 1 ) { //same as the following, just quicker.
        double r = val ( vend ); //last entry!
        if ( r < 0 ) return -inf;
        vec Psi ( nPsi + dimx );
        Psi ( 0 ) = -1.0;
        Psi.set_subvector ( 1, val ( 0, vend - 1 ) ); // fill the rest

        double Vq = V.qform ( Psi );
        return -0.5* ( nu*log ( r ) + Vq / r );
    } else {
        mat Tmp;
        if (nPsi>0) {
            mat Th = reshape ( val ( 0, nPsi * dimx - 1 ), nPsi, dimx );
            Tmp = concat_vertical ( -eye ( dimx ), Th );
        } else {
            Tmp = -eye(dimx);
        }
        fsqmat R ( reshape ( val ( nPsi*dimx, vend ), dimx, dimx ) );
        double ldetR = R.logdet();
        if ( !std::isfinite(ldetR) ) return -inf;
        fsqmat iR ( dimx );
        R.inv ( iR );

        return -0.5* ( nu*ldetR + trace ( iR.to_mat() *Tmp.T() *V.to_mat() *Tmp ) );
    }
}

double egiw::lognc() const {
    const vec& D = V._D();

    double m = nu - nPsi - dimx - 1;
#define    log2  0.693147180559945286226763983
#define    logpi 1.144729885849400163877476189
#define log2pi 1.83787706640935
#define Inf std::numeric_limits<double>::infinity()

    double nkG = 0.5 * dimx * ( -nPsi * log2pi + sum ( log ( D.mid ( dimx, nPsi ) ) ) );
    // temporary for lgamma in Wishart
    double lg = 0;
    for ( int i = 0; i < dimx; i++ ) {
        lg += lgamma ( 0.5 * ( m - i ) );
    }

    double nkW = 0.5 * ( m * sum ( log ( D ( 0, dimx - 1 ) ) ) ) \
                 - 0.5 * dimx * ( m * log2 + 0.5 * ( dimx - 1 ) * log2pi )  - lg;

//    bdm_assert_debug ( ( ( -nkG - nkW ) > -Inf ) && ( ( -nkG - nkW ) < Inf ), "ARX improper" );
    if ( -nkG - nkW == Inf ) {
        cout << "??" << endl;
    }
    return -nkG - nkW;
}

vec egiw::est_theta() const {
    if ( dimx == 1 ) {
        const mat &L = V._L();
        int end = L.rows() - 1;

        mat iLsub = ltuinv ( L ( dimx, end, dimx, end ) );

        vec L0 = L.get_col ( 0 );

        return iLsub * L0 ( 1, end );
    } else {
        bdm_error ( "ERROR: est_theta() not implemented for dimx>1" );
        return vec();
    }
}

void egiw::factorize ( mat &M, ldmat &Vz, ldmat &Lam ) const {
    const mat &L = V._L();
    const vec &D = V._D();
    int end = L.rows() - 1;

    Lam = ldmat ( L ( 0, dimx - 1, 0, dimx - 1 ), D ( 0, dimx - 1 ) );  //exp val of R

    if (dimx<=end) {
        Vz = ldmat ( L ( dimx, end, dimx, end ), D ( dimx, end ) );
        mat iLsub = ltuinv ( Vz._L() );
        // set mean value
        mat Lpsi = L ( dimx, end, 0, dimx - 1 );
        M = iLsub * Lpsi;
    }
    /*    if ( 0 ) { // test with Peterka
            mat VF = V.to_mat();
            mat Vf = VF ( 0, dimx - 1, 0, dimx - 1 );
            mat Vzf = VF ( dimx, end, 0, dimx - 1 );
            mat VZ = VF ( dimx, end, dimx, end );

            mat Lam2 = Vf - Vzf.T() * inv ( VZ ) * Vzf;
        }*/
}

ldmat egiw::est_theta_cov() const {
    if ( dimx == 1 ) {
        const mat &L = V._L();
        const vec &D = V._D();
        int end = D.length() - 1;

        mat Lsub = L ( 1, end, 1, end );
//        mat Dsub = diag ( D ( 1, end ) );

        ldmat LD ( inv ( Lsub ).T(), 1.0 / D ( 1, end ) );
        return LD;

    } else {
        bdm_error ( "ERROR: est_theta_cov() not implemented for dimx>1" );
        return ldmat();
    }

}

vec egiw::mean() const {

    if ( dimx == 1 ) {
        const vec &D = V._D();
        int end = D.length() - 1;

        vec m ( dim );
        m.set_subvector ( 0, est_theta() );
        m ( end ) = D ( 0 ) / ( nu - nPsi - 2 * dimx - 2 );
        return m;
    } else {
        mat M;
        mat R;
        mean_mat ( M, R );
        return concat ( cvectorize ( M ), cvectorize ( R ) );
    }

}

vec egiw::variance() const {
    int l = V.rows();
    // cut out rest of lower-right part of V
    // invert it
    ldmat itmp;
    if (dimx<l) {
        const ldmat tmp ( V, linspace ( dimx, l - 1 ) );
        tmp.inv ( itmp );
    }
    // following Wikipedia notation
    // m=nu-nPsi-dimx-1, p=dimx
    double mp1p = nu - nPsi - 2 * dimx; // m-p+1
    double mp1m = mp1p - 2;     // m-p-1

    if ( dimx == 1 ) {
        double cove = V._D() ( 0 ) / mp1m ;

        vec var ( l );
        var.set_subvector ( 0, diag ( itmp.to_mat() ) *cove );
        var ( l - 1 ) = cove * cove / ( mp1m - 2 );
        return var;
    } else {
        ldmat Vll ( V, linspace ( 0, dimx - 1 ) ); // top-left part of V
        mat Y = Vll.to_mat();
        mat varY ( Y.rows(), Y.cols() );

        double denom = ( mp1p - 1 ) * mp1m * mp1m * ( mp1m - 2 );         // (m-p)(m-p-1)^2(m-p-3)

        int i, j;
        for ( i = 0; i < Y.rows(); i++ ) {
            for ( j = 0; j < Y.cols(); j++ ) {
                varY ( i, j ) = ( mp1p * Y ( i, j ) * Y ( i, j ) + mp1m * Y ( i, i ) * Y ( j, j ) ) / denom;
            }
        }
        vec mean_dR = diag ( Y ) / mp1m; // corresponds to cove
        vec var_th = diag ( itmp.to_mat() );
        vec var_Th ( mean_dR.length() *var_th.length() );
        // diagonal of diag(mean_dR) \kron diag(var_th)
        for ( int i = 0; i < mean_dR.length(); i++ ) {
            var_Th.set_subvector ( i*var_th.length(), var_th*mean_dR ( i ) );
        }

        return concat ( var_Th, cvectorize ( varY ) );
    }
}

void egiw::mean_mat ( mat &M, mat&R ) const {
    const mat &L = V._L();
    const vec &D = V._D();
    int end = L.rows() - 1;

    ldmat ldR ( L ( 0, dimx - 1, 0, dimx - 1 ), D ( 0, dimx - 1 ) / ( nu - nPsi - 2*dimx - 2 ) ); //exp val of R

    // set mean value
    if (dimx<=end) {
        mat iLsub = ltuinv ( L ( dimx, end, dimx, end ) );
        mat Lpsi = L ( dimx, end, 0, dimx - 1 );
        M = iLsub * Lpsi;
    }
    R = ldR.to_mat()  ;
}

void egiw::log_register ( bdm::logger& L, const string& prefix ) {
    epdf::log_register ( L, prefix );
    if ( log_level[logvartheta] ) {
        int th_dim = dim - dimx*dimx; // dimension - dimension of cov
        L.add_vector( log_level, logvartheta, RV ( th_dim ), prefix );
    }
}

void egiw::log_write() const {
    epdf::log_write();
    if ( log_level[logvartheta] ) {
        mat M;
        ldmat Lam;
        ldmat Vz;
        factorize ( M, Vz, Lam );
        if( log_level[logvartheta] )
            log_level.store( logvartheta, cvectorize ( est_theta_cov().to_mat() ) );
    }
}

void egiw::from_setting ( const Setting &set ) {
    epdf::from_setting ( set );
    UI::get ( dimx, set, "dimx", UI::compulsory );
    if ( !UI::get ( nu, set, "nu", UI::optional ) ) {
        nu = -1;
    }
    mat Vful;
        if (set.exists("V.L")){
                UI::get(V, set, "V", UI::optional);
        } else {
        if ( !UI::get ( Vful, set, "fV", UI::optional ) ) {
            vec dV;
            UI::get ( dV, set, "dV", UI::compulsory );
            set_parameters ( dimx, ldmat ( dV ), nu );

        } else {
            set_parameters ( dimx, Vful, nu );
        }
    }
}

void egiw::to_setting ( Setting& set ) const {
    epdf::to_setting ( set );
    UI::save ( dimx, set, "dimx" );
    UI::save ( V, set, "V" );
    UI::save ( nu, set, "nu" );
};

void egiw::validate() {
    eEF::validate();
    nPsi = V.rows() - dimx;
    dim = dimx * ( dimx + nPsi );
        if (dim<dimx) {bdm_error("Check if matrix V has correct dimension");}
        
    if ( nu < 0 ) {
        nu = 0.1 + nPsi + 2 * dimx + 2; // +2 assures finite expected value of R
        // terms before that are sufficient for finite normalization
    }

    // check sizes, rvs etc.
    // also check if RV are meaningful!!!
    // meaningful =  rv for theta  and rv for r are split!
}
void multiBM::bayes ( const vec &yt, const vec &cond ) {
    if ( frg < 1.0 ) {
        beta *= frg;
        last_lognc = est.lognc();
    }
    beta += yt;
    if ( evalll ) {
        ll = est.lognc() - last_lognc;
    }
}

double multiBM::logpred ( const vec &yt ) const {
    eDirich pred ( est );
    vec &beta = pred._beta();

    double lll;
    if ( frg < 1.0 ) {
        beta *= frg;
        lll = pred.lognc();
    } else if ( evalll ) {
        lll = last_lognc;
    } else {
        lll = pred.lognc();
    }

    beta += yt;
    return pred.lognc() - lll;
}
void multiBM::flatten ( const BMEF* B, double weight=1.0 ) {
    const multiBM* E = dynamic_cast<const multiBM*> ( B );
    // sum(beta) should be equal to sum(B.beta)
    const vec &Eb = E->beta;//const_cast<multiBM*> ( E )->_beta();
    beta *= ( sum ( Eb ) / sum ( beta ) * weight);
    if ( evalll ) {
        last_lognc = est.lognc();
    }
}

vec egamma::sample() const {
    vec smp ( dim );
    int i;

    for ( i = 0; i < dim; i++ ) {
        if ( beta ( i ) > std::numeric_limits<double>::epsilon() ) {
            GamRNG.setup ( alpha ( i ), beta ( i ) );
        } else {
            GamRNG.setup ( alpha ( i ), std::numeric_limits<double>::epsilon() );
        }
#pragma omp critical
        smp ( i ) = GamRNG();
    }

    return smp;
}


double egamma::evallog_nn ( const vec &val ) const {
    double res = 0.0; //the rest will be added
    int i;

    if ( any ( val <= 0. ) ) return -inf;
    if ( any ( beta <= 0. ) ) return -inf;
    for ( i = 0; i < dim; i++ ) {
        res += ( alpha ( i ) - 1 ) * std::log ( val ( i ) ) - beta ( i ) * val ( i );
    }
    return res;
}

double egamma::lognc() const {
    double res = 0.0; //will be added
    int i;

    for ( i = 0; i < dim; i++ ) {
        res += lgamma ( alpha ( i ) ) - alpha ( i ) * std::log ( beta ( i ) ) ;
    }

    return res;
}

void egamma::from_setting ( const Setting &set ) {
    epdf::from_setting ( set ); // reads rv
    UI::get ( alpha, set, "alpha", UI::compulsory );
    UI::get ( beta, set, "beta", UI::compulsory );
}

void egamma::to_setting ( Setting &set ) const
{
    epdf::to_setting( set );
    UI::save( alpha, set, "alpha" );
    UI::save( beta, set, "beta" );
}


void egamma::validate() {
    eEF::validate();
    bdm_assert ( alpha.length() == beta.length(), "parameters do not match" );
    dim = alpha.length();
}

void mgamma::set_parameters ( double k0, const vec &beta0 ) {
    k = k0;
    iepdf.set_parameters ( k * ones ( beta0.length() ), beta0 );
}


void mgamma::from_setting ( const Setting &set ) {
    pdf::from_setting ( set ); // reads rv and rvc
    vec betatmp; // ugly but necessary
    UI::get ( betatmp, set, "beta", UI::compulsory );
    UI::get ( k, set, "k", UI::compulsory );
    set_parameters ( k, betatmp );
}

void mgamma::to_setting  (Setting  &set) const {
    pdf::to_setting(set);
    UI::save( _beta, set, "beta");
    UI::save( k, set, "k");

}

void mgamma::validate() {
    pdf_internal<egamma>::validate();

    dim = _beta.length();
    dimc = _beta.length();
}

void eEmp::resample ( RESAMPLING_METHOD method ) {
    ivec ind = zeros_i ( n );
    bdm::resample(w,ind,method);
    // copy the internals according to ind
    for (int i = 0; i < n; i++ ) {
        if ( ind ( i ) != i ) {
            samples ( i ) = samples ( ind ( i ) );
        }
        w ( i ) = 1.0 / n;
    }
}

void resample ( const vec &w, ivec &ind, RESAMPLING_METHOD method ) {
    int n = w.length();
    ind = zeros_i ( n );
    ivec N_babies = zeros_i ( n );
    vec cumDist = cumsum ( w );
    vec u ( n );
    int i, j, parent;
    double u0;

    switch ( method ) {
    case MULTINOMIAL:
        u ( n - 1 ) = pow ( UniRNG.sample(), 1.0 / n );

        for ( i = n - 2; i >= 0; i-- ) {
            u ( i ) = u ( i + 1 ) * pow ( UniRNG.sample(), 1.0 / ( i + 1 ) );
        }

        break;

    case STRATIFIED:

        for ( i = 0; i < n; i++ ) {
            u ( i ) = ( i + UniRNG.sample() ) / n;
        }

        break;

    case SYSTEMATIC:
        u0 = UniRNG.sample();

        for ( i = 0; i < n; i++ ) {
            u ( i ) = ( i + u0 ) / n;
        }

        break;

    default:
        bdm_error ( "PF::resample(): Unknown resampling method" );
    }

    // U is now full
    j = 0;

    for ( i = 0; i < n; i++ ) {
        while ( u ( i ) > cumDist ( j ) ) j++;

        N_babies ( j ) ++;
    }
    // We have assigned new babies for each Particle
    // Now, we fill the resulting index such that:
    // * particles with at least one baby should not move *
    // This assures that reassignment can be done inplace;

    // find the first parent;
    parent = 0;
    while ( N_babies ( parent ) == 0 ) parent++;

    // Build index
    for ( i = 0; i < n; i++ ) {
        if ( N_babies ( i ) > 0 ) {
            ind ( i ) = i;
            N_babies ( i ) --; //this index was now replicated;
        } else {
            // test if the parent has been fully replicated
            // if yes, find the next one
            while ( ( N_babies ( parent ) == 0 ) || ( N_babies ( parent ) == 1 && parent > i ) ) parent++;

            // Replicate parent
            ind ( i ) = parent;

            N_babies ( parent ) --; //this index was now replicated;
        }

    }
}

void eEmp::set_statistics ( const vec &w0, const epdf &epdf0 ) {
    dim = epdf0.dimension();
    w = w0;
    w /= sum ( w0 );//renormalize
    n = w.length();
    samples.set_size ( n );

    for ( int i = 0; i < n; i++ ) {
        samples ( i ) = epdf0.sample();
    }
}

void eEmp::set_samples ( const epdf* epdf0 ) {
    w = 1;
    w /= sum ( w );//renormalize

    for ( int i = 0; i < n; i++ ) {
        samples ( i ) = epdf0->sample();
    }
}

void migamma_ref::from_setting ( const Setting &set ) {
    migamma::from_setting(set);
    vec ref;
    double k,l;

    UI::get ( ref, set, "ref" , UI::compulsory );
    UI::get( k, set, "k", UI::compulsory );
    UI::get( l, set, "l", UI::compulsory );
    set_parameters ( k, ref, l );
}

void  migamma_ref::to_setting  (Setting  &set) const {
    migamma::to_setting(set);
    UI::save ( pow ( refl, 1/(1.0 - l) ), set, "ref");
    UI::save(l,set,"l");
    UI::save(k,set,"k");
}

void mlognorm::from_setting ( const Setting &set ) {
    pdf_internal<elognorm>::from_setting(set);
    vec mu0;
    double k;
    UI::get ( mu0, set, "mu0", UI::compulsory );
    UI::get( k, set, "k", UI::compulsory );
    set_parameters ( mu0.length(), k );
    condition ( mu0 );
}

void mlognorm::to_setting  (Setting  &set) const {
    pdf_internal<elognorm>::to_setting(set);
    UI::save ( exp(mu + sig2), set, "mu0");

    // inversion of sig2 = 0.5 * log ( k * k + 1 );
    double k = sqrt( exp( 2 * sig2 ) - 1  );
    UI::save(k,set,"k");
}


void mlstudent::condition ( const vec &cond ) {
    if ( cond.length() > 0 ) {
        iepdf._mu() = A * cond + mu_const;
    } else {
        iepdf._mu() =  mu_const;
    }
    double zeta;
    //ugly hack!
    if ( ( cond.length() + 1 ) == Lambda.rows() ) {
        zeta = Lambda.invqform ( concat ( cond, vec_1 ( 1.0 ) ) );
    } else {
        zeta = Lambda.invqform ( cond );
    }
    _R = Re;
    _R *= ( 1 + zeta );// / ( nu ); << nu is in Re!!!!!!
}

void eEmp::qbounds ( vec &lb, vec &ub, double perc ) const {
    // lb in inf so than it will be pushed below;
    lb.set_size ( dim );
    ub.set_size ( dim );
    lb = std::numeric_limits<double>::infinity();
    ub = -std::numeric_limits<double>::infinity();
    int j;
    for ( int i = 0; i < n; i++ ) {
        for ( j = 0; j < dim; j++ ) {
            if ( samples ( i ) ( j ) < lb ( j ) ) {
                lb ( j ) = samples ( i ) ( j );
            }
            if ( samples ( i ) ( j ) > ub ( j ) ) {
                ub ( j ) = samples ( i ) ( j );
            }
        }
    }
}

void eEmp::to_setting ( Setting &set ) const {
    epdf::to_setting( set );
    UI::save ( samples, set, "samples" );
    UI::save ( w, set, "w" );
}

void eEmp::from_setting ( const Setting &set ) {
    epdf::from_setting( set );

    UI::get( samples, set, "samples", UI::compulsory );
    UI::get ( w, set, "w", UI::compulsory );
}

void eEmp::validate () {
    epdf::validate();
    bdm_assert (samples.length()==w.length(),"samples and weigths are of different lengths");
    n = w.length();
    if (n>0)
        pdf::dim = samples ( 0 ).length();
}

void eDirich::from_setting ( const Setting &set ) {
    epdf::from_setting ( set );
    UI::get ( beta, set, "beta", UI::compulsory );
}

void eDirich::validate() {
    //check rv
    eEF::validate();
    dim = beta.length();
}

void eDirich::to_setting ( Setting &set ) const
{
    eEF::to_setting( set );
    UI::save( beta, set, "beta" );
}

void euni::from_setting ( const Setting &set ) {
    epdf::from_setting ( set ); // reads rv and rvc

    UI::get ( high, set, "high", UI::compulsory );
    UI::get ( low, set, "low", UI::compulsory );
    set_parameters ( low, high );

}

void     euni::to_setting  (Setting  &set) const {
    epdf::to_setting ( set );
    UI::save ( high, set, "high" );
    UI::save ( low, set, "low" );
}

void euni::validate() {
    epdf::validate();
    bdm_assert ( high.length() == low.length(), "Incompatible high and low vectors" );
    dim = high.length();
    bdm_assert ( min ( distance ) > 0.0, "bad support" );
}

void mgdirac::from_setting(const Setting& set) {
    pdf::from_setting(set);
    g=UI::build<fnc>(set,"g",UI::compulsory);
}

void mgdirac::to_setting(Setting &set) const {
    pdf::to_setting(set);
    UI::save(g.get(), set, "g");
}

void mgdirac::validate() {
    pdf::validate();
    dim = g->dimension();
    dimc = g->dimensionc();
}

void mDirich::from_setting ( const Setting &set ) {
    pdf::from_setting ( set ); // reads rv and rvc
    if ( _rv()._dsize() > 0 ) {
        rvc = _rv().copy_t ( -1 );
    }
    vec beta0;
    if ( !UI::get ( beta0, set, "beta0", UI::optional ) ) {
        beta0 = ones ( _rv()._dsize() );
    }
    if ( !UI::get ( betac, set, "betac", UI::optional ) ) {
        betac = 0.1 * ones ( _rv()._dsize() );
    }
    _beta = beta0;

    UI::get ( k, set, "k", UI::compulsory );
}

void mDirich::to_setting  (Setting  &set) const {
    pdf::to_setting(set);
    UI::save( _beta, set, "beta0");
    UI::save( betac, set, "betac");
    UI::save ( k, set, "k" );
}


void mDirich::validate() {
    pdf_internal<eDirich>::validate();
    bdm_assert ( _beta.length() == betac.length(), "beta0 and betac are not compatible" );
    if ( _rv()._dsize() > 0 ) {
        bdm_assert ( ( _rv()._dsize() == dimension() ) , "Size of rv does not match with beta" );
    }
    dimc = _beta.length();
}


void mBeta::from_setting ( const Setting &set ) {
    pdf::from_setting ( set ); // reads rv and rvc
    if ( _rv()._dsize() > 0 ) {
        rvc = _rv().copy_t ( -1 );
    }
    if ( !UI::get ( iepdf.beta, set, "beta", UI::optional ) ) {
        iepdf.beta = ones ( _rv()._dsize() );
    }
    if ( !UI::get ( iepdf.alpha, set, "alpha", UI::optional ) ) {
        iepdf.alpha = ones ( _rv()._dsize() );
    }
    if ( !UI::get ( betac, set, "betac", UI::optional ) ) {
        betac = 0.1 * ones ( _rv()._dsize() );
    }

    UI::get ( k, set, "k", UI::compulsory );
}

void mBeta::to_setting  (Setting  &set) const {
    pdf::to_setting(set);
    UI::save( iepdf.beta, set, "beta");
    UI::save( betac, set, "betac");
    UI::save ( k, set, "k" );
}

}