int main(void) {
int var, i, j, nb_variable, nb_component;
const int nb_vector = 500;
double *pweight = NULL;
bool *fparam = NULL;
int* perm;
StatError error;
const char * mixpath= "./tmp.mix", * spmixpath= "./tmp_sp.mix";
const char * gnupath = "./tmp_mix", * gnu_datapath = "./tmp_mix_data";
const char * margpath= "./marg_mix", * gnu_tmppath = "./tmp_mix_d";
const char * np_modelpath= "./np_model.mix", * gnu_tmpnppath = "./tmp_mix_d";
const char * entropy_data= "cluster_vectors.vec";
const char * output_path = "tmp_entropy_output.vec";
Distribution *marginal = NULL;
DiscreteDistributionData *marginal_histo = NULL;
Vectors *vec = NULL;
MultivariateMixture *mv1 = NULL, *mv_cp = NULL;
MultivariateMixture *mv_np1 = NULL, *mv_np_estim = NULL;
MultivariateMixture *mv_estim = NULL;
MultivariateMixtureData *mv_data = NULL, *cluster = NULL;
DiscreteParametric **dt1 = NULL, **dt2 = NULL;
DiscreteParametricProcess **ppcomponent = NULL;
// constructors of Mv_Mixture
mv1 = new MultivariateMixture();
// destructor of MultivariateMixture
delete mv1;
mv1= NULL;
nb_variable = 2;
nb_component = 3;
dt1 = new DiscreteParametric*[nb_component];
dt2 = new DiscreteParametric*[nb_component];
pweight = new double[nb_component];
ppcomponent = new DiscreteParametricProcess*[nb_variable];
pweight[0] = 0.1;
pweight[1] = 0.2;
pweight[2] = 0.7;
dt1[0] = new DiscreteParametric(0, BINOMIAL, 2, 12, D_DEFAULT, 0.1);
dt1[1] = new DiscreteParametric(0, BINOMIAL, 0, 10, D_DEFAULT, 0.5);
dt1[2] = new DiscreteParametric(0, BINOMIAL, 3, 10, D_DEFAULT, 0.8);
dt2[0] = new DiscreteParametric(0, POISSON, 2, I_DEFAULT, 8.0, D_DEFAULT);
dt2[1] = new DiscreteParametric(0, POISSON, 4, I_DEFAULT, 5.0, D_DEFAULT);
dt2[2] = new DiscreteParametric(0, POISSON, 0, I_DEFAULT, 2.0, D_DEFAULT);
cout << "Observation distributions for variable 1:" << endl;
for (i = 0; i < nb_component; i++) {
dt1[i]-> ascii_print(cout);
}
ppcomponent[0] = new DiscreteParametricProcess(nb_component, dt1);
ppcomponent[1] = new DiscreteParametricProcess(nb_component, dt2);
for (i = 0; i < nb_component; i++) {
delete dt1[i];
dt1[i] = NULL;
delete dt2[i];
dt2[i] = NULL;
}
cout << endl;
mv1 = new MultivariateMixture(nb_component, pweight, nb_variable, ppcomponent, NULL);
cout << "Mixture of " << nb_component << " components with " <<
nb_variable << " variables:" << endl;
mv1->ascii_write(cout, true);
cout << endl;
// copy
mv_cp = new MultivariateMixture(*mv1);
cout << "Copy constructor of MultivariateMixture: " << endl;
mv_cp->ascii_write(cout);
cout << endl;
// destructor of MultivariateMixture
delete mv_cp;
mv_cp= NULL;
delete mv1;
mv1= NULL;
cout << "MultivariateMixture_building (print into file " << mixpath << "): " << endl;
mv1 = multivariate_mixture_building(error , nb_component ,
nb_variable, pweight, ppcomponent, NULL);
if (mv1 == NULL)
cout << error;
else {
mv1->ascii_write(error, mixpath, true);
delete mv1;
mv1= NULL;
}
cout << endl;
cout << "Read MultivariateMixture from file " << mixpath << ": " << endl;
mv1 = multivariate_mixture_ascii_read(error , mixpath);
if (mv1 == NULL) {
cout << error;
return 1;
}
else {
mv1->ascii_write(cout, true);
}
cout << endl;
mv_data = mv1->simulation(error, nb_vector);
if (mv_data == NULL) {
cout << error;
return 1;
}
else {
mv_data->ascii_write(cout, true);
}
cout << endl;
cout << "Gnuplot output for MultivariateMixture ('" << gnupath << "' file)" << endl;
mv1->plot_write(error, gnupath, "");
cout << error << endl;
cout << "Gnuplot output for MultivariateMixture_data ('" << gnu_datapath << "' file)" << endl;
mv_data->plot_write(error, gnu_datapath, "");
cout << error << endl;
cout << "Extract marginal distribution for variable 1" << endl;
marginal = mv1->extract_distribution(error, 1);
marginal_histo = mv_data->extract_marginal(error, 1);
if (marginal != NULL) {
marginal->ascii_print(cout, false, false, false);
delete marginal;
}
else
cout << error;
if (marginal_histo != NULL) {
cout << "Gnuplot output for marginal ('" << margpath << "' file)" << endl;
marginal_histo->plot_write(error, margpath);
delete marginal_histo;
}
else
cout << error;
cout << "Estimate MultivariateMixture from initial model: " << endl;
mv_estim = mv_data->mixture_estimation(error, cout, *mv1);
if (mv_estim == NULL) {
cout << error;
}
else {
mv_estim->ascii_write(cout, true);
mv_estim->plot_write(error, gnu_tmppath, "");
mv_estim->spreadsheet_write(error, spmixpath);
cout << error << endl;
delete mv_estim;
mv_estim = NULL;
}
cout << endl;
fparam = new bool[2];
fparam[0] = true;
fparam[1] = false;
cout << "Estimate MultivariateMixture from initial nb_component: " << endl;
mv_estim = mv_data->mixture_estimation(error, cout, 3, I_DEFAULT, fparam);
delete [] fparam;
if (mv_estim == NULL) {
cout << error;
}
else {
mv_estim->ascii_write(cout, true);
mv_estim->plot_write(error, gnu_tmppath, "");
delete mv_estim;
mv_estim = NULL;
}
cout << endl;
delete mv1;
mv1= NULL;
delete mv_data;
mv_data = NULL;
for (var = 0; var < nb_variable; var++) {
delete ppcomponent[var];
ppcomponent[var] = NULL;
}
cout << "Read non parametric model: " << endl;
mv_np1 = multivariate_mixture_ascii_read(error, np_modelpath);
if (mv_np1 == NULL) {
cout << error;
return 1;
}
else {
cout << "Value: " << endl;
mv_np1->ascii_write(cout, false);
mv_np1->spreadsheet_write(error, spmixpath);
cout << error << endl;
}
// permutation of the states
perm = new int[mv_np1->get_nb_component()];
for(j= 0; j < mv_np1->get_nb_component(); j++)
perm[mv_np1->get_nb_component()-j -1]=j;
mv_np1->state_permutation(error, perm);
cout << error;
cout << "Permutation of the states: " << endl;
mv_np1->ascii_write(cout, false);
delete [] perm;
perm = NULL;
cout << endl;
mv_data = mv_np1->simulation(error, nb_vector);
if (mv_data == NULL) {
cout << error;
return 1;
}
cout << "Extract marginal distribution for variable 2" << endl;
marginal = mv_np1->extract_distribution(error, 2);
if (marginal != NULL) {
marginal->ascii_print(cout, false, false, false);
delete marginal;
}
else
cout << error;
cout << "Gnuplot output for MultivariateMixture_data ('" << gnu_tmpnppath << "' file)" << endl;
mv_data->plot_write(error, gnu_tmpnppath, "");
cout << error << endl;
cout << "Estimate MultivariateMixture from initial nb_component: " << endl;
mv_estim = mv_data->mixture_estimation(error, cout, 3);
if (mv_estim == NULL) {
cout << error;
return 1;
}
else {
mv_estim->ascii_write(cout, true);
mv_estim->plot_write(error, gnu_tmpnppath, "");
}
// permutation of the states
perm = new int[mv_estim->get_nb_component()];
for(j= 0; j < mv_estim->get_nb_component(); j++)
perm[mv_estim->get_nb_component()-j -1]=j;
mv_estim->state_permutation(error, perm);
cout << error;
cout << "Permutation of the states: " << endl;
mv_estim->ascii_write(cout, false);
delete [] perm;
perm = NULL;
cout << endl;
cout << "Cluster individuals: " << endl;
cluster = mv_estim->cluster(error, *mv_data);
if (cluster == NULL) {
cout << error;
}
else {
cout << "Estimated states: " << endl;
for (i = 0; i < 10; i++)
cout << cluster->get_int_vector(i, 0) << "\t";
cout << "..." << endl;
delete cluster;
cluster = NULL;
}
cout << "State entropy: " << endl;
cluster = mv_estim->cluster(error, *mv_data, VITERBI, true);
if (cluster == NULL) {
cout << error;
}
else {
cout << "State entropies: " << endl;
for (i = 0; i < 10; i++)
cout << cluster->get_real_vector(i, 3) << "\t";
cout << "..." << endl;
delete mv_estim;
mv_estim = NULL;
delete cluster;
cluster = NULL;
}
// compute state entropies on data
cout << "Compute state entropies on data" << endl;
vec = vectors_ascii_read(error, entropy_data);
if (vec == NULL) {
cout << error;
return 1;
}
mv_estim = vec->mixture_estimation(error, cout, 3, 300);
if (mv_estim == NULL) {
cout << error;
return 1;
}
else
mv_estim->ascii_write(cout, true);
cluster = mv_estim->cluster(error, *vec, VITERBI, true);
if (cluster == NULL) {
cout << error;
}
else {
cluster->ascii_write(error, output_path);
}
delete mv_estim;
delete cluster;
delete vec;
delete mv_data;
delete mv_np1;
delete [] dt1;
delete [] dt2;
delete [] pweight;
delete [] ppcomponent;
return 0;
}
