// currently only hacked for spheres, with radius and sd as two parameters
bool HipGISAXS::fit_steepest_descent(real_t zcut,
real_t radius_min, real_t radius_max, real_t radius_num,
real_t sd_min, real_t sd_max, real_t sd_num,
unsigned int dim, MPI::Intracomm& world_comm,
int x_min, int x_max, int x_step) {
int mpi_rank = world_comm.Get_rank();
if(!init_steepest_fit(world_comm, zcut)) return false;
int num_alphai = 0, num_phi = 0, num_tilt = 0;;
real_t alphai_min, alphai_max, alphai_step;
HiGInput::instance().scattering_alphai(alphai_min, alphai_max, alphai_step);
if(alphai_max < alphai_min) alphai_max = alphai_min;
if(alphai_min == alphai_max || alphai_step == 0) num_alphai = 1;
else num_alphai = (alphai_max - alphai_min) / alphai_step + 1;
real_t phi_min, phi_max, phi_step;
HiGInput::instance().scattering_inplanerot(phi_min, phi_max, phi_step);
if(phi_step == 0) num_phi = 1;
else num_phi = (phi_max - phi_min) / phi_step + 1;
real_t tilt_min, tilt_max, tilt_step;
HiGInput::instance().scattering_tilt(tilt_min, tilt_max, tilt_step);
if(tilt_step == 0) num_tilt = 1;
else num_tilt = (tilt_max - tilt_min) / tilt_step + 1;
std::cout << "** Num alphai: " << num_alphai << std::endl
<< "** Num phi: " << num_phi << std::endl
<< "** Num tilt: " << num_tilt << std::endl;
// prepare parameters
std::vector<std::vector<real_t> > params;
int num_params = 2;
std::vector<real_t> temp;
real_t deltap = 0.0;
if(radius_num <= 1)
temp.push_back(radius_min);
else {
deltap = fabs(radius_max - radius_min) / (radius_num - 1);
for(int i = 0; i < radius_num; ++ i) {
temp.push_back(radius_min + i * deltap);
} // for
} // if-else
params.push_back(temp);
temp.clear();
if(sd_num <= 1)
temp.push_back(sd_min);
else {
deltap = fabs(sd_max - sd_min) / (sd_num - 1);
for(int i = 0; i < sd_num; ++ i) {
temp.push_back(sd_min + i * deltap);
} // for
} // if-else
params.push_back(temp);
temp.clear();
// this will work only on one shape and one structure
const real_t err_threshold = 1e-8;
const unsigned int max_iter = 200;
std::vector<real_t> param_vals;
//param_vals.push_back(16.0);
//param_vals.push_back(6.0);
param_vals.push_back(23.0);
param_vals.push_back(2.0);
std::vector<real_t> param_deltas;
param_deltas.push_back(0.05);
param_deltas.push_back(0.05);
real_t gamma_const = 0.05;
real_t qdeltay = QGrid::instance().delta_y();
real_t alpha_i = alphai_min;
// high level of parallelism here (alphai, phi, tilt) for dynamicity ...
for(int i = 0; i < num_alphai; i ++, alpha_i += alphai_step) {
real_t alphai = alpha_i * PI_ / 180;
real_t phi = phi_min;
for(int j = 0; j < num_phi; j ++, phi += phi_step) {
real_t tilt = tilt_min;
for(int k = 0; k < num_tilt; k ++, tilt += tilt_step) {
std::cout << "-- Computing reference GISAXS "
<< i * num_phi * num_tilt + j * num_tilt + k + 1 << " / "
<< num_alphai * num_phi * num_tilt
<< " [alphai = " << alpha_i << ", phi = " << phi
<< ", tilt = " << tilt << "] ..." << std::endl;
/* run the reference gisaxs simulation using input params */
real_t* ref_data = NULL;
if(!run_gisaxs(alpha_i, alphai, phi, tilt, ref_data, world_comm)) {
if(mpi_rank == 0) std::cerr << "error: could not finish successfully" << std::endl;
return false;
} // if
if(dim != 1) {
std::cerr << "uh-oh: only 1D is supported for now" << std::endl;
return false;
} // if
real_t* ref_z_cut = new (std::nothrow) real_t[nqy_];
for(unsigned int iy = 0; iy < nqy_; ++ iy) {
// assuming nqz_ == 1 ...
ref_z_cut[iy] = ref_data[nqx_ * iy + 0];
} // for
delete[] ref_data;
// this will store z cut values for each iteration for plotting later
real_t* z_cuts = new (std::nothrow) real_t[nqy_ * max_iter];
real_t* temp_zcuts = new (std::nothrow) real_t[nqy_];
// do some preprocessing
// start the main loop, bound by max_iter and err_threshold
// compute gisaxs for current parameter values
// compute the neighbors parameter values
// for 12 combinations of current and neighbors, compute gisaxs and error
// compute the derivatives (gradient) and error stuff
// update parameter values
// compute the error surface
real_t err = 10.0;
std::vector<real_t> param1_list;
std::vector<real_t> param2_list;
structure_iterator_t structure_iter = HiGInput::instance().structure_begin();
Structure* structure = &((*structure_iter).second);
Shape* shape = HiGInput::instance().shape(*structure);
shape_param_iterator_t shape_param = (*shape).param_begin();
real_t* data = NULL;
std::vector<real_t> param_error_data;
for(unsigned int iter = 0; iter < max_iter; ++ iter) {
param1_list.clear();
param1_list.push_back(param_vals[0] - 2 * param_deltas[0]); // p1mm
param1_list.push_back(param_vals[0] - param_deltas[0]); // p1m
param1_list.push_back(param_vals[0]); // p1
param1_list.push_back(param_vals[0] + param_deltas[0]); // p1p
param1_list.push_back(param_vals[0] + 2 * param_deltas[0]); // p1pp
param2_list.clear();
param2_list.push_back(param_vals[1] - 2 * param_deltas[1]); // p2mm
param2_list.push_back(param_vals[1] - param_deltas[1]); // p2m
param2_list.push_back(param_vals[1]); // p2
param2_list.push_back(param_vals[1] + param_deltas[1]); // p2p
param2_list.push_back(param_vals[1] + 2 * param_deltas[1]); // p2pp
// current point
(*shape_param).second.mean(param1_list[2]);
(*shape_param).second.deviation(param2_list[2]);
if(!run_gisaxs(alpha_i, alphai, phi, tilt, data, world_comm)) {
if(mpi_rank == 0)
std::cerr << "error: could not finish successfully" << std::endl;
return false;
} // if
for(unsigned int iy = 0; iy < nqy_; ++ iy) {
// assuming nqz_ == 1 ...
z_cuts[iter * nqy_ + iy] = data[nqx_ * iy];
} // for
delete[] data; data = NULL;
real_t err22 = compute_cut_fit_error(z_cuts + iter * nqy_, ref_z_cut, qdeltay);
// 12 neighbors
(*shape_param).second.mean(param1_list[0]);
(*shape_param).second.deviation(param2_list[2]);
if(!run_gisaxs(alpha_i, alphai, phi, tilt, data, world_comm)) {
if(mpi_rank == 0)
std::cerr << "error: could not finish successfully" << std::endl;
return false;
} // if
for(unsigned int iy = 0; iy < nqy_; ++ iy) {
// assuming nqz_ == 1 ...
temp_zcuts[iy] = data[nqx_ * iy];
} // for
delete[] data; data = NULL;
real_t err02 = compute_cut_fit_error(temp_zcuts, ref_z_cut, qdeltay);
(*shape_param).second.mean(param1_list[1]);
(*shape_param).second.deviation(param2_list[1]);
if(!run_gisaxs(alpha_i, alphai, phi, tilt, data, world_comm)) {
if(mpi_rank == 0)
std::cerr << "error: could not finish successfully" << std::endl;
return false;
} // if
for(unsigned int iy = 0; iy < nqy_; ++ iy) {
// assuming nqz_ == 1 ...
temp_zcuts[iy] = data[nqx_ * iy];
} // for
delete[] data; data = NULL;
real_t err11 = compute_cut_fit_error(temp_zcuts, ref_z_cut, qdeltay);
(*shape_param).second.mean(param1_list[1]);
(*shape_param).second.deviation(param2_list[2]);
if(!run_gisaxs(alpha_i, alphai, phi, tilt, data, world_comm)) {
if(mpi_rank == 0)
std::cerr << "error: could not finish successfully" << std::endl;
return false;
} // if
for(unsigned int iy = 0; iy < nqy_; ++ iy) {
// assuming nqz_ == 1 ...
temp_zcuts[iy] = data[nqx_ * iy];
} // for
delete[] data; data = NULL;
real_t err12 = compute_cut_fit_error(temp_zcuts, ref_z_cut, qdeltay);
(*shape_param).second.mean(param1_list[1]);
(*shape_param).second.deviation(param2_list[3]);
if(!run_gisaxs(alpha_i, alphai, phi, tilt, data, world_comm)) {
if(mpi_rank == 0)
std::cerr << "error: could not finish successfully" << std::endl;
return false;
} // if
for(unsigned int iy = 0; iy < nqy_; ++ iy) {
// assuming nqz_ == 1 ...
temp_zcuts[iy] = data[nqx_ * iy];
} // for
delete[] data; data = NULL;
real_t err13 = compute_cut_fit_error(temp_zcuts, ref_z_cut, qdeltay);
(*shape_param).second.mean(param1_list[2]);
(*shape_param).second.deviation(param2_list[0]);
if(!run_gisaxs(alpha_i, alphai, phi, tilt, data, world_comm)) {
if(mpi_rank == 0)
std::cerr << "error: could not finish successfully" << std::endl;
return false;
} // if
for(unsigned int iy = 0; iy < nqy_; ++ iy) {
// assuming nqz_ == 1 ...
temp_zcuts[iy] = data[nqx_ * iy];
} // for
delete[] data; data = NULL;
real_t err20 = compute_cut_fit_error(temp_zcuts, ref_z_cut, qdeltay);
(*shape_param).second.mean(param1_list[2]);
(*shape_param).second.deviation(param2_list[1]);
if(!run_gisaxs(alpha_i, alphai, phi, tilt, data, world_comm)) {
if(mpi_rank == 0)
std::cerr << "error: could not finish successfully" << std::endl;
return false;
} // if
for(unsigned int iy = 0; iy < nqy_; ++ iy) {
// assuming nqz_ == 1 ...
temp_zcuts[iy] = data[nqx_ * iy];
} // for
delete[] data; data = NULL;
real_t err21 = compute_cut_fit_error(temp_zcuts, ref_z_cut, qdeltay);
(*shape_param).second.mean(param1_list[2]);
(*shape_param).second.deviation(param2_list[3]);
if(!run_gisaxs(alpha_i, alphai, phi, tilt, data, world_comm)) {
if(mpi_rank == 0)
std::cerr << "error: could not finish successfully" << std::endl;
return false;
} // if
for(unsigned int iy = 0; iy < nqy_; ++ iy) {
// assuming nqz_ == 1 ...
temp_zcuts[iy] = data[nqx_ * iy];
} // for
delete[] data; data = NULL;
real_t err23 = compute_cut_fit_error(temp_zcuts, ref_z_cut, qdeltay);
(*shape_param).second.mean(param1_list[2]);
(*shape_param).second.deviation(param2_list[4]);
if(!run_gisaxs(alpha_i, alphai, phi, tilt, data, world_comm)) {
if(mpi_rank == 0)
std::cerr << "error: could not finish successfully" << std::endl;
return false;
} // if
for(unsigned int iy = 0; iy < nqy_; ++ iy) {
// assuming nqz_ == 1 ...
temp_zcuts[iy] = data[nqx_ * iy];
} // for
delete[] data; data = NULL;
real_t err24 = compute_cut_fit_error(temp_zcuts, ref_z_cut, qdeltay);
(*shape_param).second.mean(param1_list[3]);
(*shape_param).second.deviation(param2_list[1]);
if(!run_gisaxs(alpha_i, alphai, phi, tilt, data, world_comm)) {
if(mpi_rank == 0)
std::cerr << "error: could not finish successfully" << std::endl;
return false;
} // if
for(unsigned int iy = 0; iy < nqy_; ++ iy) {
// assuming nqz_ == 1 ...
temp_zcuts[iy] = data[nqx_ * iy];
} // for
delete[] data; data = NULL;
real_t err31 = compute_cut_fit_error(temp_zcuts, ref_z_cut, qdeltay);
(*shape_param).second.mean(param1_list[3]);
(*shape_param).second.deviation(param2_list[2]);
if(!run_gisaxs(alpha_i, alphai, phi, tilt, data, world_comm)) {
if(mpi_rank == 0)
std::cerr << "error: could not finish successfully" << std::endl;
return false;
} // if
for(unsigned int iy = 0; iy < nqy_; ++ iy) {
// assuming nqz_ == 1 ...
temp_zcuts[iy] = data[nqx_ * iy];
} // for
delete[] data; data = NULL;
real_t err32 = compute_cut_fit_error(temp_zcuts, ref_z_cut, qdeltay);
(*shape_param).second.mean(param1_list[3]);
(*shape_param).second.deviation(param2_list[3]);
if(!run_gisaxs(alpha_i, alphai, phi, tilt, data, world_comm)) {
if(mpi_rank == 0)
std::cerr << "error: could not finish successfully" << std::endl;
return false;
} // if
for(unsigned int iy = 0; iy < nqy_; ++ iy) {
// assuming nqz_ == 1 ...
temp_zcuts[iy] = data[nqx_ * iy];
} // for
delete[] data; data = NULL;
real_t err33 = compute_cut_fit_error(temp_zcuts, ref_z_cut, qdeltay);
(*shape_param).second.mean(param1_list[4]);
(*shape_param).second.deviation(param2_list[2]);
if(!run_gisaxs(alpha_i, alphai, phi, tilt, data, world_comm)) {
if(mpi_rank == 0)
std::cerr << "error: could not finish successfully" << std::endl;
return false;
} // if
for(unsigned int iy = 0; iy < nqy_; ++ iy) {
// assuming nqz_ == 1 ...
temp_zcuts[iy] = data[nqx_ * iy];
} // for
delete[] data; data = NULL;
real_t err42 = compute_cut_fit_error(temp_zcuts, ref_z_cut, qdeltay);
// 22 0
// 02 1mm
// 11 1m2m
// 12 1m
// 13 1m2p
// 20 2mm
// 21 2m
// 23 2p
// 24 2pp
// 31 1p2m
// 32 1p
// 33 1p2p
// 42 1pp
real_t derr1 = (err32 - err12) / (2 * param_deltas[0]);
real_t derr2 = (err23 - err21) / (2 * param_deltas[1]);
err = sqrt(derr1 * derr1 + derr2 * derr2);
std::cout << "++ Iteration: " << iter << ", Error: " << err << std::endl;
std::cout << "++ Parameter 1: " << param_vals[0]
<< ", Parameter 2: " << param_vals[1] << std::endl;
param_error_data.push_back(iter);
param_error_data.push_back(param_vals[0]);
param_error_data.push_back(param_vals[1]);
param_error_data.push_back(err);
if(err < err_threshold) break;
real_t herr11 = (err42 + err02 - 2 * err22) /
(4 * param_deltas[0] * param_deltas[0]);
real_t herr12 = (err33 - err13 - (err31 - err11)) /
(4 * param_deltas[0] * param_deltas[1]);
real_t herr21 = (err33 - err13 - (err31 - err11)) /
(4 * param_deltas[0] * param_deltas[1]);
real_t herr22 = (err24 + err20 - 2 * err22) /
(4 * param_deltas[1] * param_deltas[1]);
real_t* herr = new (std::nothrow) real_t[2 * 2];
herr[0] = herr11;
herr[1] = herr12;
herr[2] = herr21;
herr[3] = herr22;
real_t* herrinv;
mldivide(2, herr, herrinv);
param_vals[0] -= gamma_const * (herrinv[0] * derr1 + herrinv[1] * derr2);
param_vals[1] -= gamma_const * (herrinv[2] * derr1 + herrinv[3] * derr2);
delete[] herrinv;
delete[] herr;
} // for
// compute the error surface
std::vector<std::vector<real_t> >::iterator mean_iter = params.begin();
std::vector<std::vector<real_t> >::iterator sd_iter = mean_iter + 1;
std::vector<real_t> err_surface;
for(std::vector<real_t>::iterator curr_mean = (*mean_iter).begin();
curr_mean != (*mean_iter).end(); ++ curr_mean) {
for(std::vector<real_t>::iterator curr_sd = (*sd_iter).begin();
curr_sd != (*sd_iter).end(); ++ curr_sd) {
(*shape_param).second.mean(*curr_mean);
(*shape_param).second.deviation(*curr_sd);
if(!run_gisaxs(alpha_i, alphai, phi, tilt, data, world_comm)) {
if(mpi_rank == 0)
std::cerr << "error: could not finish successfully" << std::endl;
return false;
} // if
for(unsigned int iy = 0; iy < nqy_; ++ iy) {
// assuming nqz_ == 1 ...
temp_zcuts[iy] = data[nqx_ * iy];
} // for
delete[] data; data = NULL;
real_t curr_err = compute_cut_fit_error(temp_zcuts, ref_z_cut, qdeltay);
err_surface.push_back(*curr_mean);
err_surface.push_back(*curr_sd);
err_surface.push_back(curr_err);
} // for
} // for
// write data to files
// define output filename
std::stringstream alphai_b, phi_b, tilt_b;
std::string alphai_s, phi_s, tilt_s;
alphai_b << alpha_i; alphai_s = alphai_b.str();
phi_b << phi; phi_s = phi_b.str();
tilt_b << tilt; tilt_s = tilt_b.str();
std::string param_error_file(HiGInput::instance().param_pathprefix() +
"/" + HiGInput::instance().runname() +
"/param_error_ai=" + alphai_s + "_rot=" + phi_s +
"_tilt=" + tilt_s + ".dat");
std::string z_cut_file(HiGInput::instance().param_pathprefix() +
"/" + HiGInput::instance().runname() +
"/z_cut_ai=" + alphai_s + "_rot=" + phi_s +
"_tilt=" + tilt_s + ".dat");
std::string err_surf_file(HiGInput::instance().param_pathprefix() +
"/" + HiGInput::instance().runname() +
"/err_surf_ai=" + alphai_s + "_rot=" + phi_s +
"_tilt=" + tilt_s + ".dat");
// write param_error_data
std::ofstream param_error_f(param_error_file.c_str());
for(std::vector<real_t>::iterator pei = param_error_data.begin();
pei != param_error_data.end(); pei += 4) {
param_error_f << *pei << "\t" << *(pei + 1) << "\t" << *(pei + 2) << "\t"
<< *(pei + 3) << std::endl;
} // for
param_error_f.close();
// write ref_z_cut and z_cuts
std::ofstream zcut_f(z_cut_file.c_str());
for(unsigned int yy = 0; yy < nqy_; ++ yy) {
zcut_f << ref_z_cut[yy] << "\t";
} // for
zcut_f << std::endl;
for(unsigned int i = 0; i < max_iter; ++ i) {
for(unsigned int yy = 0; yy < nqy_; ++ yy) {
zcut_f << z_cuts[i * nqy_ + yy] << "\t";
} // for
zcut_f << std::endl;
} // for
zcut_f.close();
// write error surface
std::ofstream err_surf_f(err_surf_file.c_str());
for(std::vector<real_t>::iterator surfi = err_surface.begin();
surfi != err_surface.end(); surfi += 3) {
err_surf_f << *surfi << "\t" << *(surfi + 1) << "\t" << *(surfi + 2) << std::endl;
} // for
err_surf_f.close();
(*shape_param).second.mean(22.0);
(*shape_param).second.deviation(7.0);
param_error_data.clear();
delete[] temp_zcuts;
delete[] z_cuts;
delete[] ref_z_cut;
std::cout << "parameter values: " << param_vals[0] << ", " << param_vals[1]
<< " [error: " << err << "]" << std::endl;
// synchronize all procs after each run
world_comm.Barrier();
} // for tilt
} // for phi
} // for alphai
return true;
} // HipGISAXS::fit_all_gisaxs()
//
// % Erased all parts not wanted for codegeneration /Martin %%
// Arguments : emxArray_real_T *s
// emxArray_boolean_T *outliers
// emxArray_boolean_T *outliers2
// Return Type : void
//
void locateOutliers(emxArray_real_T *s, emxArray_boolean_T *outliers, emxArray_boolean_T *outliers2)
{
int ia;
int ic;
emxArray_real_T *b_y1;
int orderForDim;
int iyLead;
double work_data_idx_0;
int m;
double tmp1;
double tmp2;
emxArray_real_T *SigmaMat;
emxArray_int32_T *r0;
emxArray_int32_T *r1;
emxArray_real_T *a;
unsigned int s_idx_0;
emxArray_real_T *C;
int br;
double mu;
emxArray_real_T *b_s;
/* locateOutliers: locates artifacts/outliers from data series */
/* */
/* Inputs: s = array containg data series */
/* method = artifact removal method to use. */
/* methods: 'percent' = percentage filter: locates data > x percent diff than previous data point. */
/* 'sd' = standard deviation filter: locates data > x stdev away from mean. */
/* 'above' = Threshold filter: locates data > threshold value */
/* 'below' = Threshold filter: locates data < threshold value */
/* 'median' = median filter. Outliers are located. */
/* Outputs: outliers = logical array of whether s is artifact/outlier or not */
/* eg. - [0 0 0 1 0], 1=artifact, 0=normal */
/* */
/* Examples: */
/* Locate outliers with 20% percentage filter: */
/* outliers = locateOutlers(s,'percent',0.2) */
/* Locate outliers that are above a threshold of 0.5: */
/* outliers = locateOutlers(s,'thresh','above',0.5) */
/* Locate outliers with median filter: */
/* outliers = locateOutlers(s,'median',4,5) */
/* */
/* %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% */
/* Copyright (C) 2010, John T. Ramshur, [email protected] */
/* */
/* This file is part of HRVAS */
/* */
/* HRVAS is free software: you can redistribute it and/or modify */
/* it under the terms of the GNU General Public License as published by */
/* the Free Software Foundation, either version 3 of the License, or */
/* (at your option) any later version. */
/* */
/* HRVAS is distributed in the hope that it will be useful, */
/* but WITHOUT ANY WARRANTY; without even the implied warranty of */
/* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the */
/* GNU General Public License for more details. */
/* */
/* You should have received a copy of the GNU General Public License */
/* along with Foobar. If not, see <http://www.gnu.org/licenses/>. */
/* %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% */
ia = outliers->size[0];
outliers->size[0] = s->size[0];
emxEnsureCapacity((emxArray__common *)outliers, ia, (int)sizeof(boolean_T));
ic = s->size[0];
for (ia = 0; ia < ic; ia++) {
outliers->data[ia] = false;
}
/* preallocate */
if (1 > s->size[0] - 1) {
ic = 0;
} else {
ic = s->size[0] - 1;
}
emxInit_real_T(&b_y1, 1);
if (s->size[0] == 0) {
ia = b_y1->size[0];
b_y1->size[0] = 0;
emxEnsureCapacity((emxArray__common *)b_y1, ia, (int)sizeof(double));
} else {
if (s->size[0] - 1 <= 1) {
orderForDim = s->size[0] - 1;
} else {
orderForDim = 1;
}
if (orderForDim < 1) {
ia = b_y1->size[0];
b_y1->size[0] = 0;
emxEnsureCapacity((emxArray__common *)b_y1, ia, (int)sizeof(double));
} else {
orderForDim = s->size[0] - 1;
ia = b_y1->size[0];
b_y1->size[0] = orderForDim;
emxEnsureCapacity((emxArray__common *)b_y1, ia, (int)sizeof(double));
if (!(b_y1->size[0] == 0)) {
orderForDim = 1;
iyLead = 0;
work_data_idx_0 = s->data[0];
for (m = 2; m <= s->size[0]; m++) {
tmp1 = s->data[orderForDim];
tmp2 = work_data_idx_0;
work_data_idx_0 = tmp1;
tmp1 -= tmp2;
orderForDim++;
b_y1->data[iyLead] = tmp1;
iyLead++;
}
}
}
}
emxInit_real_T(&SigmaMat, 1);
b_abs(b_y1, SigmaMat);
/* percent chage from previous */
/* find index of values where pChange > perLimit */
orderForDim = s->size[0];
if (2 > orderForDim) {
ia = 0;
orderForDim = 0;
} else {
ia = 1;
orderForDim = s->size[0];
}
emxInit_int32_T(&r0, 2);
iyLead = r0->size[0] * r0->size[1];
r0->size[0] = 1;
r0->size[1] = orderForDim - ia;
emxEnsureCapacity((emxArray__common *)r0, iyLead, (int)sizeof(int));
orderForDim -= ia;
for (iyLead = 0; iyLead < orderForDim; iyLead++) {
r0->data[r0->size[0] * iyLead] = ia + iyLead;
}
emxInit_int32_T1(&r1, 1);
ia = r1->size[0];
r1->size[0] = ic;
emxEnsureCapacity((emxArray__common *)r1, ia, (int)sizeof(int));
for (ia = 0; ia < ic; ia++) {
r1->data[ia] = 1 + ia;
}
ic = r0->size[0] * r0->size[1];
for (ia = 0; ia < ic; ia++) {
outliers->data[r0->data[ia]] = (SigmaMat->data[ia] / s->data[r1->data[ia] - 1] > 0.2);
}
emxFree_int32_T(&r1);
emxFree_int32_T(&r0);
emxInit_real_T1(&a, 2);
/* Reference: */
/* Clifford, G. (2002). "Characterizing Artefact in the Normal */
/* Human 24-Hour RR Time Series to Aid Identification and Artificial */
/* Replication of Circadian Variations in Human Beat to Beat Heart Rate */
/* using a Simple Threshold." */
/* */
/* Aubert, A. E., D. Ramaekers, et al. (1999). "The analysis of heart */
/* rate variability in unrestrained rats. Validation of method and */
/* results." Comput Methods Programs Biomed 60(3): 197-213. */
/* convert to logical array */
/* preallocate */
iyLead = s->size[0];
s_idx_0 = (unsigned int)s->size[0];
ia = a->size[0] * a->size[1];
a->size[0] = (int)s_idx_0;
a->size[1] = 2;
emxEnsureCapacity((emxArray__common *)a, ia, (int)sizeof(double));
for (orderForDim = 1; orderForDim <= iyLead; orderForDim++) {
a->data[orderForDim - 1] = (double)orderForDim / (double)iyLead;
a->data[(orderForDim + a->size[0]) - 1] = 1.0;
}
mldivide(a, s, b_y1);
emxInit_real_T(&C, 1);
if (b_y1->size[0] == 1) {
ia = C->size[0];
C->size[0] = a->size[0];
emxEnsureCapacity((emxArray__common *)C, ia, (int)sizeof(double));
ic = a->size[0];
for (ia = 0; ia < ic; ia++) {
C->data[ia] = 0.0;
for (iyLead = 0; iyLead < 2; iyLead++) {
C->data[ia] += a->data[ia + a->size[0] * iyLead] * b_y1->data[iyLead];
}
}
} else {
s_idx_0 = (unsigned int)a->size[0];
ia = C->size[0];
C->size[0] = (int)s_idx_0;
emxEnsureCapacity((emxArray__common *)C, ia, (int)sizeof(double));
m = a->size[0];
orderForDim = C->size[0];
ia = C->size[0];
C->size[0] = orderForDim;
emxEnsureCapacity((emxArray__common *)C, ia, (int)sizeof(double));
for (ia = 0; ia < orderForDim; ia++) {
C->data[ia] = 0.0;
}
if (a->size[0] == 0) {
} else {
orderForDim = 0;
while ((m > 0) && (orderForDim <= 0)) {
for (ic = 1; ic <= m; ic++) {
C->data[ic - 1] = 0.0;
}
orderForDim = m;
}
br = 0;
orderForDim = 0;
while ((m > 0) && (orderForDim <= 0)) {
orderForDim = 0;
for (iyLead = br; iyLead + 1 <= br + 2; iyLead++) {
if (b_y1->data[iyLead] != 0.0) {
ia = orderForDim;
for (ic = 0; ic + 1 <= m; ic++) {
ia++;
C->data[ic] += b_y1->data[iyLead] * a->data[ia - 1];
}
}
orderForDim += m;
}
br += 2;
orderForDim = m;
}
}
}
emxFree_real_T(&a);
ia = s->size[0];
emxEnsureCapacity((emxArray__common *)s, ia, (int)sizeof(double));
ic = s->size[0];
for (ia = 0; ia < ic; ia++) {
s->data[ia] -= C->data[ia];
}
emxFree_real_T(&C);
mu = mean(s);
/* mean */
orderForDim = s->size[0];
if (s->size[0] > 1) {
iyLead = s->size[0] - 1;
} else {
iyLead = s->size[0];
}
if (s->size[0] == 0) {
tmp2 = 0.0;
} else {
ia = 0;
work_data_idx_0 = s->data[0];
for (ic = 2; ic <= orderForDim; ic++) {
ia++;
work_data_idx_0 += s->data[ia];
}
work_data_idx_0 /= (double)s->size[0];
ia = 0;
tmp1 = s->data[0] - work_data_idx_0;
tmp2 = tmp1 * tmp1;
for (ic = 2; ic <= orderForDim; ic++) {
ia++;
tmp1 = s->data[ia] - work_data_idx_0;
tmp2 += tmp1 * tmp1;
}
tmp2 /= (double)iyLead;
}
emxInit_real_T(&b_s, 1);
/* standard deviation */
/* Create a matrix of mean values by replicating the mu vector for n rows */
repmat(mu, (double)s->size[0], b_y1);
/* Create a matrix of standard deviation values by replicating the sigma vector for n rows */
repmat(std::sqrt(tmp2), (double)s->size[0], SigmaMat);
/* Create a matrix of zeros and ones, where ones indicate the location of outliers */
ia = b_s->size[0];
b_s->size[0] = s->size[0];
emxEnsureCapacity((emxArray__common *)b_s, ia, (int)sizeof(double));
ic = s->size[0];
for (ia = 0; ia < ic; ia++) {
b_s->data[ia] = s->data[ia] - b_y1->data[ia];
}
b_abs(b_s, b_y1);
ia = outliers2->size[0];
outliers2->size[0] = b_y1->size[0];
emxEnsureCapacity((emxArray__common *)outliers2, ia, (int)sizeof(boolean_T));
ic = b_y1->size[0];
emxFree_real_T(&b_s);
for (ia = 0; ia < ic; ia++) {
outliers2->data[ia] = (b_y1->data[ia] > 3.0 * SigmaMat->data[ia]);
}
emxFree_real_T(&b_y1);
emxFree_real_T(&SigmaMat);
/* Reference: */
/* Aubert, A. E., D. Ramaekers, et al. (1999). "The analysis of heart */
/* rate variability in unrestrained rats. Validation of method and */
/* results." Comput Methods Programs Biomed 60(3): 197-213. */
/* convert to logical array */
}
Matrix& LASSO::train(Matrix& X, Matrix& Y, Options& options) {
int p = size(X, 2);
int ny = size(Y, 2);
double epsilon = options.epsilon;
int maxIter = options.maxIter;
double lambda = options.lambda;
bool calc_OV = options.calc_OV;
bool verbose = options.verbose;
/*XNX = [X, -X];
H_G = XNX' * XNX;
D = repmat(diag(H_G), [1, n_y]);
XNXTY = XNX' * Y;
A = (X' * X + lambda * eye(p)) \ (X' * Y);*/
Matrix& XNX = horzcat(2, &X, &uminus(X));
Matrix& H_G = XNX.transpose().mtimes(XNX);
double* Q = new double[size(H_G, 1)];
for (int i = 0; i < size(H_G, 1); i++) {
Q[i] = H_G.getEntry(i, i);
}
Matrix& XNXTY = XNX.transpose().mtimes(Y);
Matrix& A = mldivide(
plus(X.transpose().mtimes(X), times(lambda, eye(p))),
X.transpose().mtimes(Y)
);
/*AA = [subplus(A); subplus(-A)];
C = -XNXTY + lambda;
Grad = C + H_G * AA;
tol = epsilon * norm(Grad);
PGrad = zeros(size(Grad));*/
Matrix& AA = vertcat(2, &subplus(A), &subplus(uminus(A)));
Matrix& C = plus(uminus(XNXTY), lambda);
Matrix& Grad = plus(C, mtimes(H_G, AA));
double tol = epsilon * norm(Grad);
Matrix& PGrad = zeros(size(Grad));
std::list<double> J;
double fval = 0;
// J(1) = sum(sum((Y - X * A).^2)) / 2 + lambda * sum(sum(abs(A)));
if (calc_OV) {
fval = sum(sum(pow(minus(Y, mtimes(X, A)), 2))) / 2 +
lambda * sum(sum(abs(A)));
J.push_back(fval);
}
Matrix& I_k = Grad.copy();
double d = 0;
int k = 0;
DenseVector& SFPlusCi = *new DenseVector(AA.getColumnDimension());
Matrix& S = H_G;
Vector** SRows = null;
if (typeid(H_G) == typeid(DenseMatrix))
SRows = denseMatrix2DenseRowVectors(S);
else
SRows = sparseMatrix2SparseRowVectors(S);
Vector** CRows = null;
if (typeid(C) == typeid(DenseMatrix))
CRows = denseMatrix2DenseRowVectors(C);
else
CRows = sparseMatrix2SparseRowVectors(C);
double** FData = ((DenseMatrix&) AA).getData();
double* FRow = null;
double* pr = null;
int K = 2 * p;
while (true) {
/*I_k = Grad < 0 | AA > 0;
I_k_com = not(I_k);
PGrad(I_k) = Grad(I_k);
PGrad(I_k_com) = 0;*/
_or(I_k, lt(Grad, 0), gt(AA, 0));
Matrix& I_k_com = _not(I_k);
assign(PGrad, Grad);
logicalIndexingAssignment(PGrad, I_k_com, 0);
d = norm(PGrad, inf);
if (d < tol) {
if (verbose)
println("Converge successfully!");
break;
}
/*for i = 1:2*p
AA(i, :) = max(AA(i, :) - (C(i, :) + H_G(i, :) * AA) ./ (D(i, :)), 0);
end
A = AA(1:p,:) - AA(p+1:end,:);*/
for (int i = 0; i < K; i++) {
// SFPlusCi = SRows[i].operate(AA);
operate(SFPlusCi, *SRows[i], AA);
plusAssign(SFPlusCi, *CRows[i]);
timesAssign(SFPlusCi, 1 / Q[i]);
pr = SFPlusCi.getPr();
// F(i, :) = max(F(i, :) - (S(i, :) * F + C(i, :)) / D[i]), 0);
// F(i, :) = max(F(i, :) - SFPlusCi, 0)
FRow = FData[i];
for (int j = 0; j < AA.getColumnDimension(); j++) {
FRow[j] = max(FRow[j] - pr[j], 0);
}
}
// Grad = plus(C, mtimes(H_G, AA));
plus(Grad, C, mtimes(H_G, AA));
k = k + 1;
if (k > maxIter) {
if (verbose)
println("Maximal iterations");
break;
}
if (calc_OV) {
fval = sum(sum(pow(minus(Y, mtimes(XNX, AA)), 2))) / 2 +
lambda * sum(sum(abs(AA)));
J.push_back(fval);
}
if (k % 10 == 0 && verbose) {
if (calc_OV)
fprintf("Iter %d - ||PGrad||: %f, ofv: %f\n", k, d, J.back());
else
fprintf("Iter %d - ||PGrad||: %f\n", k, d);
}
}
Matrix& res = minus(
AA.getSubMatrix(0, p - 1, 0, ny - 1),
AA.getSubMatrix(p, 2 * p - 1, 0, ny - 1)
);
return res;
}