00001
00013 #ifndef EF_H
00014 #define EF_H
00015
00016 #include <itpp/itbase.h>
00017 #include "../math/libDC.h"
00018 #include "libBM.h"
00019 #include "../itpp_ext.h"
00020
00021
00022 using namespace itpp;
00023
00024
00026 extern Uniform_RNG UniRNG;
00028 extern Normal_RNG NorRNG;
00030 extern Gamma_RNG GamRNG;
00031
00038 class eEF : public epdf {
00039 public:
00040
00042 eEF ( const RV &rv ) :epdf ( rv ) {};
00044 virtual double lognc()const =0;
00046 virtual void tupdate ( double phi, mat &vbar, double nubar ) {};
00048 virtual void dupdate ( mat &v,double nu=1.0 ) {};
00049 };
00050
00057 class mEF : public mpdf {
00058
00059 public:
00061 mEF ( const RV &rv0, const RV &rvc0 ) :mpdf ( rv0,rvc0 ) {};
00062 };
00063
00069 template<class sq_T>
00070
00071 class enorm : public eEF {
00072 protected:
00074 vec mu;
00076 sq_T R;
00078 int dim;
00079 public:
00080
00082 enorm ( RV &rv );
00084 void set_parameters ( const vec &mu,const sq_T &R );
00086 void tupdate ( double phi, mat &vbar, double nubar );
00088 void dupdate ( mat &v,double nu=1.0 );
00089
00090 vec sample() const;
00092 mat sample ( int N ) const;
00093 double eval ( const vec &val ) const ;
00094 double evalpdflog ( const vec &val ) const;
00095 double lognc () const;
00096 vec mean() const {return mu;}
00097
00098
00100 vec& _mu() {return mu;}
00101
00103 void set_mu(const vec mu0) { mu=mu0;}
00104
00106 sq_T& _R() {return R;}
00107
00109 mat getR () {return R.to_mat();}
00110 };
00111
00117 class egiw : public eEF {
00118 protected:
00120 ldmat V;
00122 double nu;
00123 public:
00125 egiw(RV rv, mat V0, double nu0): eEF(rv), V(V0), nu(nu0) {
00126 it_assert_debug(rv.count()==V.rows(),"Incompatible V0.");
00127 }
00128
00129 vec sample() const;
00130 vec mean() const;
00131 double evalpdflog ( const vec &val ) const;
00132 double lognc () const;
00133
00134
00136 ldmat& _V() {return V;}
00138 double& _nu() {return nu;}
00139
00140 };
00141
00151 class egamma : public eEF {
00152 protected:
00154 vec alpha;
00156 vec beta;
00157 public :
00159 egamma ( const RV &rv ) :eEF ( rv ) {};
00161 void set_parameters ( const vec &a, const vec &b ) {alpha=a,beta=b;};
00162 vec sample() const;
00164 mat sample ( int N ) const;
00165 double evalpdflog ( const vec &val ) const;
00166 double lognc () const;
00168 void _param ( vec* &a, vec* &b ) {a=αb=β};
00169 vec mean() const {vec pom ( alpha ); pom/=beta; return pom;}
00170 };
00171
00173
00174
00175
00176
00177
00178
00180
00181
00182
00183
00184
00185
00186
00188
00189 class euni: public epdf {
00190 protected:
00192 vec low;
00194 vec high;
00196 vec distance;
00198 double nk;
00200 double lnk;
00201 public:
00203 euni ( const RV rv ) :epdf ( rv ) {}
00204 double eval ( const vec &val ) const {return nk;}
00205 double evalpdflog ( const vec &val ) const {return lnk;}
00206 vec sample() const {
00207 vec smp ( rv.count() ); UniRNG.sample_vector ( rv.count(),smp );
00208 return low+distance*smp;
00209 }
00211 void set_parameters ( const vec &low0, const vec &high0 ) {
00212 distance = high0-low0;
00213 it_assert_debug ( min ( distance ) >0.0,"bad support" );
00214 low = low0;
00215 high = high0;
00216 nk = prod ( 1.0/distance );
00217 lnk = log ( nk );
00218 }
00219 vec mean() const {vec pom=high; pom-=low; pom/=2.0; return pom;}
00220 };
00221
00222
00228 template<class sq_T>
00229 class mlnorm : public mEF {
00231 enorm<sq_T> epdf;
00232 mat A;
00233 vec& _mu;
00234 public:
00236 mlnorm ( RV &rv,RV &rvc );
00238 void set_parameters ( const mat &A, const sq_T &R );
00240 vec samplecond ( vec &cond, double &lik );
00242 mat samplecond ( vec &cond, vec &lik, int n );
00244 void condition ( vec &cond );
00245 };
00246
00256 class mgamma : public mEF {
00257 protected:
00259 egamma epdf;
00261 double k;
00263 vec* _beta;
00264
00265 public:
00267 mgamma ( const RV &rv,const RV &rvc );
00269 void set_parameters ( double k );
00271 vec samplecond ( vec &cond, double &lik );
00273 mat samplecond ( vec &cond, vec &lik, int n );
00274 void condition ( const vec &val ) {*_beta=k/val;};
00275 };
00276
00288 class mgamma_fix : public mgamma {
00289 protected:
00291 double l;
00293 vec refl;
00294 public:
00296 mgamma_fix ( const RV &rv,const RV &rvc ) : mgamma ( rv,rvc ),refl ( rv.count() ) {};
00298 void set_parameters ( double k0 , vec ref0, double l0 ) {
00299 mgamma::set_parameters ( k0 );
00300 refl=pow ( ref0,1.0-l0 );l=l0;
00301 };
00302
00303 void condition ( const vec &val ) {vec mean=elem_mult ( refl,pow ( val,l ) ); *_beta=k/mean;};
00304 };
00305
00307 enum RESAMPLING_METHOD { MULTINOMIAL = 0, STRATIFIED = 1, SYSTEMATIC = 3 };
00313 class eEmp: public epdf {
00314 protected :
00316 int n;
00318 vec w;
00320 Array<vec> samples;
00321 public:
00323 eEmp ( const RV &rv0 ,int n0 ) :epdf ( rv0 ),n ( n0 ),w ( n ),samples ( n ) {};
00325 void set_parameters ( const vec &w0, epdf* pdf0 );
00327 vec& _w() {return w;};
00329 Array<vec>& _samples() {return samples;};
00331 ivec resample ( RESAMPLING_METHOD method = SYSTEMATIC );
00333 vec sample() const {it_error ( "Not implemented" );return 0;}
00335 double evalpdflog ( const vec &val ) const {it_error ( "Not implemented" );return 0.0;}
00336 vec mean() const {
00337 vec pom=zeros ( rv.count() );
00338 for ( int i=0;i<n;i++ ) {pom+=samples ( i ) *w ( i );}
00339 return pom;
00340 }
00341 };
00342
00343
00345
00346 template<class sq_T>
00347 enorm<sq_T>::enorm ( RV &rv ) :eEF ( rv ), mu ( rv.count() ),R ( rv.count() ),dim ( rv.count() ) {};
00348
00349 template<class sq_T>
00350 void enorm<sq_T>::set_parameters ( const vec &mu0, const sq_T &R0 ) {
00351
00352 mu = mu0;
00353 R = R0;
00354 };
00355
00356 template<class sq_T>
00357 void enorm<sq_T>::dupdate ( mat &v, double nu ) {
00358
00359 };
00360
00361 template<class sq_T>
00362 void enorm<sq_T>::tupdate ( double phi, mat &vbar, double nubar ) {
00363
00364 };
00365
00366 template<class sq_T>
00367 vec enorm<sq_T>::sample() const {
00368 vec x ( dim );
00369 NorRNG.sample_vector ( dim,x );
00370 vec smp = R.sqrt_mult ( x );
00371
00372 smp += mu;
00373 return smp;
00374 };
00375
00376 template<class sq_T>
00377 mat enorm<sq_T>::sample ( int N ) const {
00378 mat X ( dim,N );
00379 vec x ( dim );
00380 vec pom;
00381 int i;
00382
00383 for ( i=0;i<N;i++ ) {
00384 NorRNG.sample_vector ( dim,x );
00385 pom = R.sqrt_mult ( x );
00386 pom +=mu;
00387 X.set_col ( i, pom );
00388 }
00389
00390 return X;
00391 };
00392
00393 template<class sq_T>
00394 double enorm<sq_T>::eval ( const vec &val ) const {
00395 double pdfl,e;
00396 pdfl = evalpdflog ( val );
00397 e = exp ( pdfl );
00398 return e;
00399 };
00400
00401 template<class sq_T>
00402 double enorm<sq_T>::evalpdflog ( const vec &val ) const {
00403
00404 return -0.5* ( +R.invqform ( mu-val ) ) - lognc();
00405 };
00406
00407 template<class sq_T>
00408 inline double enorm<sq_T>::lognc () const {
00409
00410 return -0.5* ( R.cols() * 1.83787706640935 +R.logdet());
00411 };
00412
00413 template<class sq_T>
00414 mlnorm<sq_T>::mlnorm ( RV &rv0,RV &rvc0 ) :mEF ( rv0,rvc0 ),epdf ( rv ),A ( rv0.count(),rv0.count() ),_mu(epdf._mu()) {
00415 }
00416
00417 template<class sq_T>
00418 void mlnorm<sq_T>::set_parameters ( const mat &A0, const sq_T &R0 ) {
00419 epdf.set_parameters ( zeros ( rv.count() ),R0 );
00420 A = A0;
00421 }
00422
00423 template<class sq_T>
00424 vec mlnorm<sq_T>::samplecond ( vec &cond, double &lik ) {
00425 this->condition ( cond );
00426 vec smp = epdf.sample();
00427 lik = epdf.eval ( smp );
00428 return smp;
00429 }
00430
00431 template<class sq_T>
00432 mat mlnorm<sq_T>::samplecond ( vec &cond, vec &lik, int n ) {
00433 int i;
00434 int dim = rv.count();
00435 mat Smp ( dim,n );
00436 vec smp ( dim );
00437 this->condition ( cond );
00438
00439 for ( i=0; i<n; i++ ) {
00440 smp = epdf.sample();
00441 lik ( i ) = epdf.eval ( smp );
00442 Smp.set_col ( i ,smp );
00443 }
00444
00445 return Smp;
00446 }
00447
00448 template<class sq_T>
00449 void mlnorm<sq_T>::condition ( vec &cond ) {
00450 _mu = A*cond;
00451
00452 }
00453
00455
00456
00457 #endif //EF_H