[262] | 1 | |
---|
[176] | 2 | #include "merger.h" |
---|
[384] | 3 | #include "../estim/arx.h" |
---|
[176] | 4 | |
---|
[477] | 5 | namespace bdm { |
---|
[423] | 6 | |
---|
[477] | 7 | merger_base::merger_base ( const Array<mpdf*> &S, bool own ) { |
---|
| 8 | DBG = false; |
---|
| 9 | dbg_file = NULL; |
---|
| 10 | set_sources ( S, own ); |
---|
| 11 | } |
---|
[423] | 12 | |
---|
[477] | 13 | vec merger_base::merge_points ( mat &lW ) { |
---|
| 14 | int nu = lW.rows(); |
---|
| 15 | vec result; |
---|
| 16 | ivec indW; |
---|
| 17 | bool infexist; |
---|
| 18 | switch ( METHOD ) { |
---|
| 19 | case ARITHMETIC: |
---|
| 20 | result = log ( sum ( exp ( lW ) ) ); //ugly! |
---|
| 21 | break; |
---|
| 22 | case GEOMETRIC: |
---|
| 23 | result = sum ( lW ) / nu; |
---|
| 24 | break; |
---|
| 25 | case LOGNORMAL: |
---|
| 26 | vec sumlW = sum ( lW ) ; |
---|
| 27 | indW = find ( ( sumlW < inf ) & ( sumlW > -inf ) ); |
---|
| 28 | infexist = ( indW.size() < lW.cols() ); |
---|
| 29 | vec mu; |
---|
| 30 | vec lam; |
---|
| 31 | if ( infexist ) { |
---|
| 32 | mu = sumlW ( indW ) / nu; //mean of logs |
---|
| 33 | // |
---|
| 34 | mat validlW = lW.get_cols ( indW ); |
---|
| 35 | lam = sum ( pow ( validlW - outer_product ( ones ( validlW.rows() ), mu ), 2 ) ); |
---|
| 36 | } else { |
---|
| 37 | mu = sum ( lW ) / nu; //mean of logs |
---|
| 38 | lam = sum ( pow ( lW - outer_product ( ones ( lW.rows() ), mu ), 2 ) ); |
---|
[299] | 39 | } |
---|
[477] | 40 | // |
---|
| 41 | double coef = 0.0; |
---|
| 42 | vec sq2bl = sqrt ( 2 * beta * lam ); //this term is everywhere |
---|
| 43 | switch ( nu ) { |
---|
| 44 | case 2: |
---|
| 45 | coef = ( 1 - 0.5 * sqrt ( ( 4.0 * beta - 3.0 ) / beta ) ); |
---|
| 46 | result = coef * sq2bl + mu ; |
---|
| 47 | break; |
---|
| 48 | // case 4: == can be done similar to case 2 - is it worth it??? |
---|
| 49 | default: // see accompanying document merge_lognorm_derivation.lyx |
---|
| 50 | coef = sqrt ( 1 - ( nu + 1 ) / ( 2 * beta * nu ) ); |
---|
| 51 | result = log ( besselk ( ( nu - 3 ) / 2, sq2bl * coef ) ) - log ( besselk ( ( nu - 3 ) / 2, sq2bl ) ) + mu; |
---|
| 52 | break; |
---|
[404] | 53 | } |
---|
[477] | 54 | break; |
---|
[176] | 55 | } |
---|
[477] | 56 | if ( infexist ) { |
---|
| 57 | vec tmp = -inf * ones ( lW.cols() ); |
---|
| 58 | set_subvector ( tmp, indW, result ); |
---|
| 59 | return tmp; |
---|
| 60 | } else { |
---|
| 61 | return result; |
---|
| 62 | } |
---|
| 63 | } |
---|
[176] | 64 | |
---|
[477] | 65 | void merger_mix::merge ( ) { |
---|
| 66 | Array<vec> &Smp = eSmp._samples(); //aux |
---|
| 67 | vec &w = eSmp._w(); //aux |
---|
[180] | 68 | |
---|
[477] | 69 | mat Smp_ex = ones ( dim + 1, Npoints ); // Extended samples for the ARX model - the last row is ones |
---|
| 70 | for ( int i = 0; i < Npoints; i++ ) { |
---|
| 71 | set_col_part ( Smp_ex, i, Smp ( i ) ); |
---|
| 72 | } |
---|
[180] | 73 | |
---|
[477] | 74 | if ( DBG ) *dbg_file << Name ( "Smp_0" ) << Smp_ex; |
---|
[180] | 75 | |
---|
[477] | 76 | // Stuff for merging |
---|
| 77 | vec lw_src ( Npoints ); // weights of the ith source |
---|
| 78 | vec lw_mix ( Npoints ); // weights of the approximating mixture |
---|
| 79 | vec lw ( Npoints ); // tmp |
---|
| 80 | mat lW = zeros ( Nsources, Npoints ); // array of weights of all sources |
---|
| 81 | vec vec0 ( 0 ); |
---|
[176] | 82 | |
---|
[477] | 83 | //initialize importance weights |
---|
| 84 | lw_mix = 1.0; // assuming uniform grid density -- otherwise |
---|
[300] | 85 | |
---|
[477] | 86 | // Initial component in the mixture model |
---|
| 87 | mat V0 = 1e-8 * eye ( dim + 1 ); |
---|
| 88 | ARX A0; |
---|
| 89 | A0.set_statistics ( dim, V0 ); //initial guess of Mix: |
---|
[176] | 90 | |
---|
[477] | 91 | Mix.init ( &A0, Smp_ex, Ncoms ); |
---|
| 92 | //Preserve initial mixture for repetitive estimation via flattening |
---|
| 93 | MixEF Mix_init ( Mix ); |
---|
[197] | 94 | |
---|
[477] | 95 | // ============= MAIN LOOP ================== |
---|
| 96 | bool converged = false; |
---|
| 97 | int niter = 0; |
---|
| 98 | char dbg_str[100]; |
---|
[182] | 99 | |
---|
[477] | 100 | emix* Mpred = Mix.epredictor ( ); |
---|
| 101 | vec Mix_pdf ( Npoints ); |
---|
| 102 | while ( !converged ) { |
---|
| 103 | //Re-estimate Mix |
---|
| 104 | //Re-Initialize Mixture model |
---|
| 105 | Mix.flatten ( &Mix_init ); |
---|
| 106 | Mix.bayesB ( Smp_ex, w*Npoints ); |
---|
| 107 | delete Mpred; |
---|
| 108 | Mpred = Mix.epredictor ( ); // Allocation => must be deleted at the end!! |
---|
| 109 | Mpred->set_rv ( rv ); //the predictor predicts rv of this merger |
---|
[213] | 110 | |
---|
[477] | 111 | // This will be active only later in iterations!!! |
---|
| 112 | if ( 1. / sum_sqr ( w ) < effss_coef*Npoints ) { |
---|
| 113 | // Generate new samples |
---|
| 114 | eSmp.set_samples ( Mpred ); |
---|
| 115 | for ( int i = 0; i < Npoints; i++ ) { |
---|
| 116 | //////////// !!!!!!!!!!!!! |
---|
| 117 | //if ( Smp ( i ) ( 2 ) <0 ) {Smp ( i ) ( 2 ) = 0.01; } |
---|
| 118 | set_col_part ( Smp_ex, i, Smp ( i ) ); |
---|
| 119 | //Importance of the mixture |
---|
| 120 | //lw_mix ( i ) =Mix.logpred (Smp_ex.get_col(i) ); |
---|
| 121 | lw_mix ( i ) = Mpred->evallog ( Smp ( i ) ); |
---|
[204] | 122 | } |
---|
[477] | 123 | if ( DBG ) { |
---|
| 124 | cout << "Resampling =" << 1. / sum_sqr ( w ) << endl; |
---|
| 125 | cout << Mix._e()->mean() << endl; |
---|
| 126 | cout << sum ( Smp_ex, 2 ) / Npoints << endl; |
---|
| 127 | cout << Smp_ex*Smp_ex.T() / Npoints << endl; |
---|
| 128 | } |
---|
| 129 | } |
---|
| 130 | if ( DBG ) { |
---|
| 131 | sprintf ( dbg_str, "Mpred_mean%d", niter ); |
---|
| 132 | *dbg_file << Name ( dbg_str ) << Mpred->mean(); |
---|
| 133 | sprintf ( dbg_str, "Mpred_var%d", niter ); |
---|
| 134 | *dbg_file << Name ( dbg_str ) << Mpred->variance(); |
---|
[197] | 135 | |
---|
[205] | 136 | |
---|
[477] | 137 | sprintf ( dbg_str, "Mpdf%d", niter ); |
---|
| 138 | for ( int i = 0; i < Npoints; i++ ) { |
---|
| 139 | Mix_pdf ( i ) = Mix.logpred ( Smp_ex.get_col ( i ) ); |
---|
| 140 | } |
---|
| 141 | *dbg_file << Name ( dbg_str ) << Mix_pdf; |
---|
[180] | 142 | |
---|
[477] | 143 | sprintf ( dbg_str, "Smp%d", niter ); |
---|
| 144 | *dbg_file << Name ( dbg_str ) << Smp_ex; |
---|
[180] | 145 | |
---|
[477] | 146 | } |
---|
| 147 | //Importace weighting |
---|
| 148 | for ( int i = 0; i < mpdfs.length(); i++ ) { |
---|
| 149 | lw_src = 0.0; |
---|
| 150 | //======== Same RVs =========== |
---|
| 151 | //Split according to dependency in rvs |
---|
| 152 | if ( mpdfs ( i )->dimension() == dim ) { |
---|
| 153 | // no need for conditioning or marginalization |
---|
[487] | 154 | lw_src = mpdfs ( i )->evallogcond_m ( Smp , vec(0)); |
---|
[477] | 155 | } else { |
---|
| 156 | // compute likelihood of marginal on the conditional variable |
---|
| 157 | if ( mpdfs ( i )->dimensionc() > 0 ) { |
---|
| 158 | // Make marginal on rvc_i |
---|
| 159 | epdf* tmp_marg = Mpred->marginal ( mpdfs ( i )->_rvc() ); |
---|
| 160 | //compute vector of lw_src |
---|
| 161 | for ( int k = 0; k < Npoints; k++ ) { |
---|
| 162 | // Here val of tmp_marg = cond of mpdfs(i) ==> calling dls->get_cond |
---|
| 163 | lw_src ( k ) += tmp_marg->evallog ( dls ( i )->get_cond ( Smp ( k ) ) ); |
---|
| 164 | } |
---|
| 165 | delete tmp_marg; |
---|
[198] | 166 | |
---|
| 167 | // sprintf ( str,"marg%d",niter ); |
---|
[299] | 168 | // *dbg << Name ( str ) << lw_src; |
---|
[198] | 169 | |
---|
[477] | 170 | } |
---|
| 171 | // Compute likelihood of the missing variable |
---|
| 172 | if ( dim > ( mpdfs ( i )->dimension() + mpdfs ( i )->dimensionc() ) ) { |
---|
| 173 | /////////////// |
---|
| 174 | // There are variales unknown to mpdfs(i) : rvzs |
---|
| 175 | mpdf* tmp_cond = Mpred->condition ( rvzs ( i ) ); |
---|
| 176 | // Compute likelihood |
---|
| 177 | vec lw_dbg = lw_src; |
---|
| 178 | for ( int k = 0; k < Npoints; k++ ) { |
---|
| 179 | lw_src ( k ) += log ( |
---|
| 180 | tmp_cond->evallogcond ( |
---|
| 181 | zdls ( i )->pushdown ( Smp ( k ) ), |
---|
| 182 | zdls ( i )->get_cond ( Smp ( k ) ) ) ); |
---|
| 183 | if ( !std::isfinite ( lw_src ( k ) ) ) { |
---|
| 184 | lw_src ( k ) = -1e16; |
---|
| 185 | cout << "!"; |
---|
[204] | 186 | } |
---|
[182] | 187 | } |
---|
[477] | 188 | delete tmp_cond; |
---|
[182] | 189 | } |
---|
[477] | 190 | // Compute likelihood of the partial source |
---|
| 191 | for ( int k = 0; k < Npoints; k++ ) { |
---|
[487] | 192 | lw_src ( k ) += mpdfs ( i )->evallogcond ( dls ( i )->pushdown ( Smp ( k ) ), |
---|
| 193 | dls ( i )->get_cond ( Smp ( k ) )); |
---|
[477] | 194 | } |
---|
| 195 | |
---|
[299] | 196 | } |
---|
[477] | 197 | // it_assert_debug(std::isfinite(sum(lw_src)),"bad"); |
---|
| 198 | lW.set_row ( i, lw_src ); // do not divide by mix |
---|
| 199 | } |
---|
| 200 | lw = merger_base::merge_points ( lW ); //merge |
---|
[197] | 201 | |
---|
[477] | 202 | //Importance weighting |
---|
| 203 | lw -= lw_mix; // hoping that it is not numerically sensitive... |
---|
| 204 | w = exp ( lw - max ( lw ) ); |
---|
[300] | 205 | |
---|
[477] | 206 | //renormalize |
---|
| 207 | double sumw = sum ( w ); |
---|
| 208 | if ( std::isfinite ( sumw ) ) { |
---|
| 209 | w = w / sumw; |
---|
| 210 | } else { |
---|
| 211 | it_file itf ( "merg_err.it" ); |
---|
| 212 | itf << Name ( "w" ) << w; |
---|
| 213 | } |
---|
[180] | 214 | |
---|
[477] | 215 | if ( DBG ) { |
---|
| 216 | sprintf ( dbg_str, "lW%d", niter ); |
---|
| 217 | *dbg_file << Name ( dbg_str ) << lW; |
---|
| 218 | sprintf ( dbg_str, "w%d", niter ); |
---|
| 219 | *dbg_file << Name ( dbg_str ) << w; |
---|
| 220 | sprintf ( dbg_str, "lw_m%d", niter ); |
---|
| 221 | *dbg_file << Name ( dbg_str ) << lw_mix; |
---|
[204] | 222 | } |
---|
[477] | 223 | // ==== stopping rule === |
---|
| 224 | niter++; |
---|
| 225 | converged = ( niter > stop_niter ); |
---|
| 226 | } |
---|
| 227 | delete Mpred; |
---|
[299] | 228 | // cout << endl; |
---|
[205] | 229 | |
---|
[477] | 230 | } |
---|
[176] | 231 | |
---|
[477] | 232 | // DEFAULTS FOR MERGER_BASE |
---|
| 233 | const MERGER_METHOD merger_base::DFLT_METHOD = LOGNORMAL; |
---|
| 234 | const double merger_base::DFLT_beta = 1.2; |
---|
| 235 | // DEFAULTS FOR MERGER_MIX |
---|
| 236 | const int merger_mix::DFLT_Ncoms = 10; |
---|
| 237 | const double merger_mix::DFLT_effss_coef = 0.5; |
---|
[399] | 238 | |
---|
[176] | 239 | } |
---|