CMS 3D CMS Logo

HGCalImagingAlgo.cc
Go to the documentation of this file.
2 
3 // Geometry
8 
10 //
12 #include "tbb/task_arena.h"
13 #include "tbb/tbb.h"
14 
15 using namespace hgcal_clustering;
16 
18  points_.clear();
19  minpos_.clear();
20  maxpos_.clear();
21  points_.resize(2 * (maxlayer_ + 1));
22  minpos_.resize(2 * (maxlayer_ + 1), {{0.0f, 0.0f}});
23  maxpos_.resize(2 * (maxlayer_ + 1), {{0.0f, 0.0f}});
24 }
25 
27  // loop over all hits and create the Hexel structure, skip energies below ecut
28 
29  if (dependSensor_) {
30  // for each layer and wafer calculate the thresholds (sigmaNoise and energy)
31  // once
32  computeThreshold();
33  }
34 
35  std::vector<bool> firstHit(2 * (maxlayer_ + 1), true);
36  for (unsigned int i = 0; i < hits.size(); ++i) {
37  const HGCRecHit &hgrh = hits[i];
38  DetId detid = hgrh.detid();
39  unsigned int layer = rhtools_.getLayerWithOffset(detid);
40  float thickness = 0.f;
41  // set sigmaNoise default value 1 to use kappa value directly in case of
42  // sensor-independent thresholds
43  float sigmaNoise = 1.f;
44  if (dependSensor_) {
45  thickness = rhtools_.getSiThickness(detid);
46  int thickness_index = rhtools_.getSiThickIndex(detid);
47  if (thickness_index == -1)
48  thickness_index = 3;
49  double storedThreshold = thresholds_[layer - 1][thickness_index];
50  sigmaNoise = sigmaNoise_[layer - 1][thickness_index];
51 
52  if (hgrh.energy() < storedThreshold)
53  continue; // this sets the ZS threshold at ecut times the sigma noise
54  // for the sensor
55  }
56  if (!dependSensor_ && hgrh.energy() < ecut_)
57  continue;
58 
59  // map layers from positive endcap (z) to layer + maxlayer+1 to prevent
60  // mixing up hits from different sides
61  layer += int(rhtools_.zside(detid) > 0) * (maxlayer_ + 1);
62 
63  // determine whether this is a half-hexagon
64  bool isHalf = rhtools_.isHalfCell(detid);
65  const GlobalPoint position(rhtools_.getPosition(detid));
66 
67  // here's were the KDNode is passed its dims arguments - note that these are
68  // *copied* from the Hexel
69  points_[layer].emplace_back(
70  Hexel(hgrh, detid, isHalf, sigmaNoise, thickness, &rhtools_), position.x(), position.y());
71 
72  // for each layer, store the minimum and maximum x and y coordinates for the
73  // KDTreeBox boundaries
74  if (firstHit[layer]) {
75  minpos_[layer][0] = position.x();
76  minpos_[layer][1] = position.y();
77  maxpos_[layer][0] = position.x();
78  maxpos_[layer][1] = position.y();
79  firstHit[layer] = false;
80  } else {
81  minpos_[layer][0] = std::min((float)position.x(), minpos_[layer][0]);
82  minpos_[layer][1] = std::min((float)position.y(), minpos_[layer][1]);
83  maxpos_[layer][0] = std::max((float)position.x(), maxpos_[layer][0]);
84  maxpos_[layer][1] = std::max((float)position.y(), maxpos_[layer][1]);
85  }
86 
87  } // end loop hits
88 }
89 // Create a vector of Hexels associated to one cluster from a collection of
90 // HGCalRecHits - this can be used directly to make the final cluster list -
91 // this method can be invoked multiple times for the same event with different
92 // input (reset should be called between events)
94  layerClustersPerLayer_.resize(2 * maxlayer_ + 2);
95  // assign all hits in each layer to a cluster core or halo
96  tbb::this_task_arena::isolate([&] {
97  tbb::parallel_for(size_t(0), size_t(2 * maxlayer_ + 2), [&](size_t i) {
98  KDTreeBox bounds(minpos_[i][0], maxpos_[i][0], minpos_[i][1], maxpos_[i][1]);
99  KDTree hit_kdtree;
100  hit_kdtree.build(points_[i], bounds);
101 
102  unsigned int actualLayer =
103  i > maxlayer_ ? (i - (maxlayer_ + 1)) : i; // maps back from index used for KD trees to actual layer
104 
105  double maxdensity = calculateLocalDensity(points_[i], hit_kdtree, actualLayer); // also stores rho (energy
106  // density) for each point (node)
107  // calculate distance to nearest point with higher density storing
108  // distance (delta) and point's index
109  calculateDistanceToHigher(points_[i]);
110  findAndAssignClusters(points_[i], hit_kdtree, maxdensity, bounds, actualLayer, layerClustersPerLayer_[i]);
111  });
112  });
113  //Now that we have the density per point we can store it
114  for (auto const &p : points_) {
115  setDensity(p);
116  }
117 }
118 
119 std::vector<reco::BasicCluster> HGCalImagingAlgo::getClusters(bool doSharing) {
121  std::vector<std::pair<DetId, float>> thisCluster;
122  for (auto &clsOnLayer : layerClustersPerLayer_) {
123  for (unsigned int i = 0; i < clsOnLayer.size(); ++i) {
124  double energy = 0;
125  Point position;
126 
127  //Will save the maximum density hit of the cluster
128  size_t rsmax = max_index(clsOnLayer[i]);
129 
130  if (doSharing) {
131  std::vector<unsigned> seeds = findLocalMaximaInCluster(clsOnLayer[i]);
132  // sharing found seeds.size() sub-cluster seeds in cluster i
133 
134  std::vector<std::vector<double>> fractions;
135  // first pass can have noise it in
136  shareEnergy(clsOnLayer[i], seeds, fractions);
137 
138  // reset and run second pass after vetoing seeds
139  // that result in trivial clusters (less than 2 effective cells)
140 
141  for (unsigned isub = 0; isub < fractions.size(); ++isub) {
142  double effective_hits = 0.0;
143  double energy = calculateEnergyWithFraction(clsOnLayer[i], fractions[isub]);
144  Point position = calculatePositionWithFraction(clsOnLayer[i], fractions[isub]);
145 
146  for (unsigned ihit = 0; ihit < fractions[isub].size(); ++ihit) {
147  const double fraction = fractions[isub][ihit];
148  if (fraction > 1e-7) {
149  effective_hits += fraction;
150  thisCluster.emplace_back(clsOnLayer[i][ihit].data.detid, fraction);
151  }
152  }
153 
154  if (verbosity_ < pINFO) {
155  std::cout << "\t******** NEW CLUSTER (SHARING) ********" << std::endl;
156  std::cout << "\tEff. No. of cells = " << effective_hits << std::endl;
157  std::cout << "\t Energy = " << energy << std::endl;
158  std::cout << "\t Phi = " << position.phi() << std::endl;
159  std::cout << "\t Eta = " << position.eta() << std::endl;
160  std::cout << "\t*****************************" << std::endl;
161  }
162  clusters_v_.emplace_back(energy, position, caloID, thisCluster, algoId_);
163  if (!clusters_v_.empty()) {
164  clusters_v_.back().setSeed(clsOnLayer[i][rsmax].data.detid);
165  }
166  thisCluster.clear();
167  }
168  } else {
169  position = calculatePosition(clsOnLayer[i]); // energy-weighted position
170  // std::vector< KDNode >::iterator it;
171  for (auto &it : clsOnLayer[i]) {
172  energy += it.data.isHalo ? 0. : it.data.weight;
173  // use fraction to store whether this is a Halo hit or not
174  thisCluster.emplace_back(it.data.detid, (it.data.isHalo ? 0.f : 1.f));
175  }
176  if (verbosity_ < pINFO) {
177  std::cout << "******** NEW CLUSTER (HGCIA) ********" << std::endl;
178  std::cout << "Index " << i << std::endl;
179  std::cout << "No. of cells = " << clsOnLayer[i].size() << std::endl;
180  std::cout << " Energy = " << energy << std::endl;
181  std::cout << " Phi = " << position.phi() << std::endl;
182  std::cout << " Eta = " << position.eta() << std::endl;
183  std::cout << "*****************************" << std::endl;
184  }
185  clusters_v_.emplace_back(energy, position, caloID, thisCluster, algoId_);
186  if (!clusters_v_.empty()) {
187  clusters_v_.back().setSeed(clsOnLayer[i][rsmax].data.detid);
188  }
189  thisCluster.clear();
190  }
191  }
192  }
193  return clusters_v_;
194 }
195 
197  float total_weight = 0.f;
198  float x = 0.f;
199  float y = 0.f;
200 
201  unsigned int v_size = v.size();
202  unsigned int maxEnergyIndex = 0;
203  float maxEnergyValue = 0;
204 
205  // loop over hits in cluster candidate
206  // determining the maximum energy hit
207  for (unsigned int i = 0; i < v_size; i++) {
208  if (v[i].data.weight > maxEnergyValue) {
209  maxEnergyValue = v[i].data.weight;
210  maxEnergyIndex = i;
211  }
212  }
213 
214  // Si cell or Scintillator. Used to set approach and parameters
215  int thick = rhtools_.getSiThickIndex(v[maxEnergyIndex].data.detid);
216 
217  // for hits within positionDeltaRho_c_ from maximum energy hit
218  // build up weight for energy-weighted position
219  // and save corresponding hits indices
220  std::vector<unsigned int> innerIndices;
221  for (unsigned int i = 0; i < v_size; i++) {
222  if (thick == -1 || distance2(v[i].data, v[maxEnergyIndex].data) < positionDeltaRho_c_[thick]) {
223  innerIndices.push_back(i);
224 
225  float rhEnergy = v[i].data.weight;
226  total_weight += rhEnergy;
227  // just fill x, y for scintillator
228  // for Si it is overwritten later anyway
229  if (thick == -1) {
230  x += v[i].data.x * rhEnergy;
231  y += v[i].data.y * rhEnergy;
232  }
233  }
234  }
235  // just loop on reduced vector of interesting indices
236  // to compute log weighting
237  if (thick != -1 && total_weight != 0.) { // Silicon case
238  float total_weight_log = 0.f;
239  float x_log = 0.f;
240  float y_log = 0.f;
241  for (auto idx : innerIndices) {
242  float rhEnergy = v[idx].data.weight;
243  if (rhEnergy == 0.)
244  continue;
245  float Wi = std::max(thresholdW0_[thick] + log(rhEnergy / total_weight), 0.);
246  x_log += v[idx].data.x * Wi;
247  y_log += v[idx].data.y * Wi;
248  total_weight_log += Wi;
249  }
250  total_weight = total_weight_log;
251  x = x_log;
252  y = y_log;
253  }
254 
255  if (total_weight != 0.) {
256  auto inv_tot_weight = 1. / total_weight;
257  return math::XYZPoint(x * inv_tot_weight, y * inv_tot_weight, v[maxEnergyIndex].data.z);
258  }
259  return math::XYZPoint(0, 0, 0);
260 }
261 
262 double HGCalImagingAlgo::calculateLocalDensity(std::vector<KDNode> &nd, KDTree &lp, const unsigned int layer) const {
263  double maxdensity = 0.;
264  float delta_c; // maximum search distance (critical distance) for local
265  // density calculation
266  if (layer <= lastLayerEE_)
267  delta_c = vecDeltas_[0];
268  else if (layer < firstLayerBH_)
269  delta_c = vecDeltas_[1];
270  else
271  delta_c = vecDeltas_[2];
272 
273  // for each node calculate local density rho and store it
274  for (unsigned int i = 0; i < nd.size(); ++i) {
275  // speec up search by looking within +/- delta_c window only
276  KDTreeBox search_box(
277  nd[i].dims[0] - delta_c, nd[i].dims[0] + delta_c, nd[i].dims[1] - delta_c, nd[i].dims[1] + delta_c);
278  std::vector<Hexel> found;
279  lp.search(search_box, found);
280  const unsigned int found_size = found.size();
281  for (unsigned int j = 0; j < found_size; j++) {
282  if (distance(nd[i].data, found[j]) < delta_c) {
283  nd[i].data.rho += found[j].weight;
284  maxdensity = std::max(maxdensity, nd[i].data.rho);
285  }
286  } // end loop found
287  } // end loop nodes
288  return maxdensity;
289 }
290 
291 double HGCalImagingAlgo::calculateDistanceToHigher(std::vector<KDNode> &nd) const {
292  // sort vector of Hexels by decreasing local density
293  std::vector<size_t> rs = sorted_indices(nd);
294 
295  double maxdensity = 0.0;
296  int nearestHigher = -1;
297 
298  if (!rs.empty())
299  maxdensity = nd[rs[0]].data.rho;
300  else
301  return maxdensity; // there are no hits
302  double dist2 = 0.;
303  // start by setting delta for the highest density hit to
304  // the most distant hit - this is a convention
305 
306  for (auto &j : nd) {
307  double tmp = distance2(nd[rs[0]].data, j.data);
308  if (tmp > dist2)
309  dist2 = tmp;
310  }
311  nd[rs[0]].data.delta = std::sqrt(dist2);
312  nd[rs[0]].data.nearestHigher = nearestHigher;
313 
314  // now we save the largest distance as a starting point
315  const double max_dist2 = dist2;
316  const unsigned int nd_size = nd.size();
317 
318  for (unsigned int oi = 1; oi < nd_size; ++oi) { // start from second-highest density
319  dist2 = max_dist2;
320  unsigned int i = rs[oi];
321  // we only need to check up to oi since hits
322  // are ordered by decreasing density
323  // and all points coming BEFORE oi are guaranteed to have higher rho
324  // and the ones AFTER to have lower rho
325  for (unsigned int oj = 0; oj < oi; ++oj) {
326  unsigned int j = rs[oj];
327  // Limit the search box
328  if ((nd[i].data.x - nd[j].data.x) * (nd[i].data.x - nd[j].data.x) > dist2)
329  continue;
330  if ((nd[i].data.y - nd[j].data.y) * (nd[i].data.y - nd[j].data.y) > dist2)
331  continue;
332  double tmp = distance2(nd[i].data, nd[j].data);
333  if (tmp <= dist2) { // this "<=" instead of "<" addresses the (rare) case
334  // when there are only two hits
335  dist2 = tmp;
336  nearestHigher = j;
337  }
338  }
339  nd[i].data.delta = std::sqrt(dist2);
340  nd[i].data.nearestHigher = nearestHigher; // this uses the original unsorted hitlist
341  }
342  return maxdensity;
343 }
344 int HGCalImagingAlgo::findAndAssignClusters(std::vector<KDNode> &nd,
345  KDTree &lp,
346  double maxdensity,
347  KDTreeBox<2> &bounds,
348  const unsigned int layer,
349  std::vector<std::vector<KDNode>> &clustersOnLayer) const {
350  // this is called once per layer and endcap...
351  // so when filling the cluster temporary vector of Hexels we resize each time
352  // by the number of clusters found. This is always equal to the number of
353  // cluster centers...
354 
355  unsigned int nClustersOnLayer = 0;
356  float delta_c; // critical distance
357  if (layer <= lastLayerEE_)
358  delta_c = vecDeltas_[0];
359  else if (layer < firstLayerBH_)
360  delta_c = vecDeltas_[1];
361  else
362  delta_c = vecDeltas_[2];
363 
364  std::vector<size_t> rs = sorted_indices(nd); // indices sorted by decreasing rho
365  std::vector<size_t> ds = sort_by_delta(nd); // sort in decreasing distance to higher
366 
367  const unsigned int nd_size = nd.size();
368  for (unsigned int i = 0; i < nd_size; ++i) {
369  if (nd[ds[i]].data.delta < delta_c)
370  break; // no more cluster centers to be looked at
371  if (dependSensor_) {
372  float rho_c = kappa_ * nd[ds[i]].data.sigmaNoise;
373  if (nd[ds[i]].data.rho < rho_c)
374  continue; // set equal to kappa times noise threshold
375 
376  } else if (nd[ds[i]].data.rho * kappa_ < maxdensity)
377  continue;
378 
379  nd[ds[i]].data.clusterIndex = nClustersOnLayer;
380  if (verbosity_ < pINFO) {
381  std::cout << "Adding new cluster with index " << nClustersOnLayer << std::endl;
382  std::cout << "Cluster center is hit " << ds[i] << std::endl;
383  }
384  nClustersOnLayer++;
385  }
386 
387  // at this point nClustersOnLayer is equal to the number of cluster centers -
388  // if it is zero we are done
389  if (nClustersOnLayer == 0)
390  return nClustersOnLayer;
391 
392  // assign remaining points to clusters, using the nearestHigher set from
393  // previous step (always set except
394  // for top density hit that is skipped...)
395  for (unsigned int oi = 1; oi < nd_size; ++oi) {
396  unsigned int i = rs[oi];
397  int ci = nd[i].data.clusterIndex;
398  if (ci == -1) { // clusterIndex is initialised with -1 if not yet used in cluster
399  nd[i].data.clusterIndex = nd[nd[i].data.nearestHigher].data.clusterIndex;
400  }
401  }
402 
403  // make room in the temporary cluster vector for the additional clusterIndex
404  // clusters
405  // from this layer
406  if (verbosity_ < pINFO) {
407  std::cout << "resizing cluster vector by " << nClustersOnLayer << std::endl;
408  }
409  clustersOnLayer.resize(nClustersOnLayer);
410 
411  // assign points closer than dc to other clusters to border region
412  // and find critical border density
413  std::vector<double> rho_b(nClustersOnLayer, 0.);
414  lp.clear();
415  lp.build(nd, bounds);
416  // now loop on all hits again :( and check: if there are hits from another
417  // cluster within d_c -> flag as border hit
418  for (unsigned int i = 0; i < nd_size; ++i) {
419  int ci = nd[i].data.clusterIndex;
420  bool flag_isolated = true;
421  if (ci != -1) {
422  KDTreeBox search_box(
423  nd[i].dims[0] - delta_c, nd[i].dims[0] + delta_c, nd[i].dims[1] - delta_c, nd[i].dims[1] + delta_c);
424  std::vector<Hexel> found;
425  lp.search(search_box, found);
426 
427  const unsigned int found_size = found.size();
428  for (unsigned int j = 0; j < found_size; j++) { // start from 0 here instead of 1
429  // check if the hit is not within d_c of another cluster
430  if (found[j].clusterIndex != -1) {
431  float dist = distance(found[j], nd[i].data);
432  if (dist < delta_c && found[j].clusterIndex != ci) {
433  // in which case we assign it to the border
434  nd[i].data.isBorder = true;
435  break;
436  }
437  // because we are using two different containers, we have to make sure
438  // that we don't unflag the
439  // hit when it finds *itself* closer than delta_c
440  if (dist < delta_c && dist != 0. && found[j].clusterIndex == ci) {
441  // in this case it is not an isolated hit
442  // the dist!=0 is because the hit being looked at is also inside the
443  // search box and at dist==0
444  flag_isolated = false;
445  }
446  }
447  }
448  if (flag_isolated)
449  nd[i].data.isBorder = true; // the hit is more than delta_c from any of its brethren
450  }
451  // check if this border hit has density larger than the current rho_b and
452  // update
453  if (nd[i].data.isBorder && rho_b[ci] < nd[i].data.rho)
454  rho_b[ci] = nd[i].data.rho;
455  } // end loop all hits
456 
457  // flag points in cluster with density < rho_b as halo points, then fill the
458  // cluster vector
459  for (unsigned int i = 0; i < nd_size; ++i) {
460  int ci = nd[i].data.clusterIndex;
461  if (ci != -1) {
462  if (nd[i].data.rho <= rho_b[ci])
463  nd[i].data.isHalo = true;
464  clustersOnLayer[ci].push_back(nd[i]);
465  if (verbosity_ < pINFO) {
466  std::cout << "Pushing hit " << i << " into cluster with index " << ci << std::endl;
467  }
468  }
469  }
470 
471  // prepare the offset for the next layer if there is one
472  if (verbosity_ < pINFO) {
473  std::cout << "moving cluster offset by " << nClustersOnLayer << std::endl;
474  }
475  return nClustersOnLayer;
476 }
477 
478 // find local maxima within delta_c, marking the indices in the cluster
479 std::vector<unsigned> HGCalImagingAlgo::findLocalMaximaInCluster(const std::vector<KDNode> &cluster) {
480  std::vector<unsigned> result;
481  std::vector<bool> seed(cluster.size(), true);
482  float delta_c = 2.;
483 
484  for (unsigned i = 0; i < cluster.size(); ++i) {
485  for (unsigned j = 0; j < cluster.size(); ++j) {
486  if (i != j and distance(cluster[i].data, cluster[j].data) < delta_c) {
487  if (cluster[i].data.weight < cluster[j].data.weight) {
488  seed[i] = false;
489  break;
490  }
491  }
492  }
493  }
494 
495  for (unsigned i = 0; i < cluster.size(); ++i) {
496  if (seed[i] && cluster[i].data.weight > 5e-4) {
497  // seed at i with energy cluster[i].weight
498  result.push_back(i);
499  }
500  }
501 
502  // Found result.size() sub-clusters in input cluster of length cluster.size()
503 
504  return result;
505 }
506 
508  const std::vector<double> &fractions) {
509  double norm(0.0), x(0.0), y(0.0), z(0.0);
510  for (unsigned i = 0; i < hits.size(); ++i) {
511  const double weight = fractions[i] * hits[i].data.weight;
512  norm += weight;
513  x += weight * hits[i].data.x;
514  y += weight * hits[i].data.y;
515  z += weight * hits[i].data.z;
516  }
517  math::XYZPoint result(x, y, z);
518  result /= norm;
519  return result;
520 }
521 
522 double HGCalImagingAlgo::calculateEnergyWithFraction(const std::vector<KDNode> &hits,
523  const std::vector<double> &fractions) {
524  double result = 0.0;
525  for (unsigned i = 0; i < hits.size(); ++i) {
526  result += fractions[i] * hits[i].data.weight;
527  }
528  return result;
529 }
530 
531 void HGCalImagingAlgo::shareEnergy(const std::vector<KDNode> &incluster,
532  const std::vector<unsigned> &seeds,
533  std::vector<std::vector<double>> &outclusters) {
534  std::vector<bool> isaseed(incluster.size(), false);
535  outclusters.clear();
536  outclusters.resize(seeds.size());
537  std::vector<Point> centroids(seeds.size());
538  std::vector<double> energies(seeds.size());
539 
540  if (seeds.size() == 1) { // short circuit the case of a lone cluster
541  outclusters[0].clear();
542  outclusters[0].resize(incluster.size(), 1.0);
543  return;
544  }
545 
546  // saving seeds
547 
548  // create quick seed lookup
549  for (unsigned i = 0; i < seeds.size(); ++i) {
550  isaseed[seeds[i]] = true;
551  }
552 
553  // initialize clusters to be shared
554  // centroids start off at seed positions
555  // seeds always have fraction 1.0, to stabilize fit
556  // initializing fit
557  for (unsigned i = 0; i < seeds.size(); ++i) {
558  outclusters[i].resize(incluster.size(), 0.0);
559  for (unsigned j = 0; j < incluster.size(); ++j) {
560  if (j == seeds[i]) {
561  outclusters[i][j] = 1.0;
562  centroids[i] = math::XYZPoint(incluster[j].data.x, incluster[j].data.y, incluster[j].data.z);
563  energies[i] = incluster[j].data.weight;
564  }
565  }
566  }
567 
568  // run the fit while we are less than max iterations, and clusters are still
569  // moving
570  const double minFracTot = 1e-20;
571  unsigned iter = 0;
572  const unsigned iterMax = 50;
574  const double stoppingTolerance = 1e-8;
575  const auto numberOfSeeds = seeds.size();
576  auto toleranceScaling = numberOfSeeds > 2 ? (numberOfSeeds - 1) * (numberOfSeeds - 1) : 1;
577  std::vector<Point> prevCentroids;
578  std::vector<double> frac(numberOfSeeds), dist2(numberOfSeeds);
579  while (iter++ < iterMax && diff > stoppingTolerance * toleranceScaling) {
580  for (unsigned i = 0; i < incluster.size(); ++i) {
581  const Hexel &ihit = incluster[i].data;
582  double fracTot(0.0);
583  for (unsigned j = 0; j < numberOfSeeds; ++j) {
584  double fraction = 0.0;
585  double d2 = (std::pow(ihit.x - centroids[j].x(), 2.0) + std::pow(ihit.y - centroids[j].y(), 2.0) +
586  std::pow(ihit.z - centroids[j].z(), 2.0)) /
587  sigma2_;
588  dist2[j] = d2;
589  // now we set the fractions up based on hit type
590  if (i == seeds[j]) { // this cluster's seed
591  fraction = 1.0;
592  } else if (isaseed[i]) {
593  fraction = 0.0;
594  } else {
595  fraction = energies[j] * std::exp(-0.5 * d2);
596  }
597  fracTot += fraction;
598  frac[j] = fraction;
599  }
600  // now that we have calculated all fractions for all hits
601  // assign the new fractions
602  for (unsigned j = 0; j < numberOfSeeds; ++j) {
603  if (fracTot > minFracTot || (i == seeds[j] && fracTot > 0.0)) {
604  outclusters[j][i] = frac[j] / fracTot;
605  } else {
606  outclusters[j][i] = 0.0;
607  }
608  }
609  }
610 
611  // save previous centroids
612  prevCentroids = std::move(centroids);
613  // finally update the position of the centroids from the last iteration
614  centroids.resize(numberOfSeeds);
615  double diff2 = 0.0;
616  for (unsigned i = 0; i < numberOfSeeds; ++i) {
617  centroids[i] = calculatePositionWithFraction(incluster, outclusters[i]);
618  energies[i] = calculateEnergyWithFraction(incluster, outclusters[i]);
619  // calculate convergence parameters
620  const double delta2 = (prevCentroids[i] - centroids[i]).perp2();
621  diff2 = std::max(delta2, diff2);
622  }
623  // update convergance parameter outside loop
624  diff = std::sqrt(diff2);
625  }
626 }
627 
629  // To support the TDR geometry and also the post-TDR one (v9 onwards), we
630  // need to change the logic of the vectors containing signal to noise and
631  // thresholds. The first 3 indices will keep on addressing the different
632  // thicknesses of the Silicon detectors, while the last one, number 3 (the
633  // fourth) will address the Scintillators. This change will support both
634  // geometries at the same time.
635 
636  if (initialized_)
637  return; // only need to calculate thresholds once
638 
639  initialized_ = true;
640 
641  std::vector<double> dummy;
642  const unsigned maxNumberOfThickIndices = 3;
643  dummy.resize(maxNumberOfThickIndices + 1, 0); // +1 to accomodate for the Scintillators
644  thresholds_.resize(maxlayer_, dummy);
645  sigmaNoise_.resize(maxlayer_, dummy);
646 
647  for (unsigned ilayer = 1; ilayer <= maxlayer_; ++ilayer) {
648  for (unsigned ithick = 0; ithick < maxNumberOfThickIndices; ++ithick) {
649  float sigmaNoise = 0.001f * fcPerEle_ * nonAgedNoises_[ithick] * dEdXweights_[ilayer] /
650  (fcPerMip_[ithick] * thicknessCorrection_[ithick]);
651  thresholds_[ilayer - 1][ithick] = sigmaNoise * ecut_;
652  sigmaNoise_[ilayer - 1][ithick] = sigmaNoise;
653  }
654  float scintillators_sigmaNoise = 0.001f * noiseMip_ * dEdXweights_[ilayer];
655  thresholds_[ilayer - 1][maxNumberOfThickIndices] = ecut_ * scintillators_sigmaNoise;
656  sigmaNoise_[ilayer - 1][maxNumberOfThickIndices] = scintillators_sigmaNoise;
657  }
658 }
659 
660 void HGCalImagingAlgo::setDensity(const std::vector<KDNode> &nd) {
661  // for each node calculate local density rho and store it
662  for (auto &i : nd) {
663  density_[i.data.detid] = i.data.rho;
664  } // end loop nodes
665 }
666 
667 //Density
668 Density HGCalImagingAlgo::getDensity() { return density_; }
HLT_FULL_cff.minFracTot
minFracTot
Definition: HLT_FULL_cff.py:13214
HGCalImagingAlgo::calculateEnergyWithFraction
double calculateEnergyWithFraction(const std::vector< KDNode > &, const std::vector< double > &)
Definition: HGCalImagingAlgo.cc:522
change_name.diff
diff
Definition: change_name.py:13
mps_fire.i
i
Definition: mps_fire.py:428
PositionCalc.h
hgcal_clustering::sorted_indices
std::vector< size_t > sorted_indices(const std::vector< T > &v)
Definition: HGCalClusteringAlgoBase.h:18
CaloRecHit::energy
constexpr float energy() const
Definition: CaloRecHit.h:29
hfClusterShapes_cfi.hits
hits
Definition: hfClusterShapes_cfi.py:5
reco::CaloID::DET_HGCAL_ENDCAP
Definition: CaloID.h:30
min
T min(T a, T b)
Definition: MathUtil.h:58
HLTEgPhaseIITestSequence_cff.doSharing
doSharing
Definition: HLTEgPhaseIITestSequence_cff.py:1317
mps_merge.weight
weight
Definition: mps_merge.py:88
AlCaHLTBitMon_ParallelJobs.p
p
Definition: AlCaHLTBitMon_ParallelJobs.py:153
HGCalImagingAlgo::Hexel::z
double z
Definition: HGCalImagingAlgo.h:159
gather_cfg.cout
cout
Definition: gather_cfg.py:144
CaloID.h
edm::SortedCollection
Definition: SortedCollection.h:49
HGCalImagingAlgo::Hexel
Definition: HGCalImagingAlgo.h:156
findQualityFiles.v
v
Definition: findQualityFiles.py:179
HGCalImagingAlgo::findLocalMaximaInCluster
std::vector< unsigned > findLocalMaximaInCluster(const std::vector< KDNode > &)
Definition: HGCalImagingAlgo.cc:479
CaloRecHit::detid
constexpr const DetId & detid() const
Definition: CaloRecHit.h:33
createJobs.tmp
tmp
align.sh
Definition: createJobs.py:716
newFWLiteAna.found
found
Definition: newFWLiteAna.py:118
HGCalImagingAlgo::getClusters
std::vector< reco::BasicCluster > getClusters(bool) override
Definition: HGCalImagingAlgo.cc:119
heavyIonCSV_trainingSettings.idx
idx
Definition: heavyIonCSV_trainingSettings.py:5
fileCollector.seed
seed
Definition: fileCollector.py:127
DetId
Definition: DetId.h:17
DivergingColor.frac
float frac
Definition: DivergingColor.py:175
KDTreeLinkerAlgo::search
void search(const KDTreeBox< DIM > &searchBox, std::vector< DATA > &resRecHitList)
Definition: KDTreeLinkerAlgo.h:203
HGCalImagingAlgo::getEventSetupPerAlgorithm
void getEventSetupPerAlgorithm(const edm::EventSetup &es) override
Definition: HGCalImagingAlgo.cc:17
HGCalImagingAlgo::calculatePositionWithFraction
math::XYZPoint calculatePositionWithFraction(const std::vector< KDNode > &, const std::vector< double > &)
Definition: HGCalImagingAlgo.cc:507
HLT_FULL_cff.fraction
fraction
Definition: HLT_FULL_cff.py:52806
Calorimetry_cff.thickness
thickness
Definition: Calorimetry_cff.py:115
hgcal_clustering::Density
std::map< DetId, float > Density
Definition: HGCalClusteringAlgoBase.h:43
mathSSE::sqrt
T sqrt(T t)
Definition: SSEVec.h:19
KDTreeBox
Definition: KDTreeLinkerAlgo.h:14
KDTreeLinkerAlgo
Definition: KDTreeLinkerAlgo.h:102
HGCalImagingAlgo::makeClusters
void makeClusters() override
Definition: HGCalImagingAlgo.cc:93
HCALHighEnergyHPDFilter_cfi.energy
energy
Definition: HCALHighEnergyHPDFilter_cfi.py:5
hgcal_clustering
Definition: HGCalClusteringAlgoBase.h:16
Point3DBase< float, GlobalTag >
phase1PixelTopology::layer
constexpr std::array< uint8_t, layerIndexSize > layer
Definition: phase1PixelTopology.h:99
HGCRecHit
Definition: HGCRecHit.h:14
InitialStep_cff.seeds
seeds
Definition: InitialStep_cff.py:231
HGCalImagingAlgo::findAndAssignClusters
int findAndAssignClusters(std::vector< KDNode > &, KDTree &, double, KDTreeBox< 2 > &, const unsigned int, std::vector< std::vector< KDNode >> &) const
Definition: HGCalImagingAlgo.cc:344
Point
Structure Point Contains parameters of Gaussian fits to DMRs.
Definition: DMRtrends.cc:57
CaloSubdetectorGeometry.h
math::XYZPoint
XYZPointD XYZPoint
point in space with cartesian internal representation
Definition: Point3D.h:12
HGCalImagingAlgo::Hexel::y
double y
Definition: HGCalImagingAlgo.h:158
SiStripPI::max
Definition: SiStripPayloadInspectorHelper.h:169
HGCalImagingAlgo::calculateDistanceToHigher
double calculateDistanceToHigher(std::vector< KDNode > &) const
Definition: HGCalImagingAlgo.cc:291
HGCalImagingAlgo::getDensity
Density getDensity() override
Definition: HGCalImagingAlgo.cc:668
position
static int position[264][3]
Definition: ReadPGInfo.cc:289
HGCalImagingAlgo::setDensity
void setDensity(const std::vector< KDNode > &nd)
Definition: HGCalImagingAlgo.cc:660
createfilelist.int
int
Definition: createfilelist.py:10
trackerHitRTTI::vector
Definition: trackerHitRTTI.h:21
IdealGeometryRecord.h
edm::EventSetup
Definition: EventSetup.h:58
HcalSubdetector.h
perp2
T perp2() const
Squared magnitude of transverse component.
Definition: Basic3DVectorLD.h:130
cms::cuda::for
for(int i=first, nt=offsets[nh];i< nt;i+=gridDim.x *blockDim.x)
Definition: HistoContainer.h:27
reco::CaloID
Definition: CaloID.h:17
HGCalImagingAlgo::Hexel::x
double x
Definition: HGCalImagingAlgo.h:157
HGCalImagingAlgo::computeThreshold
void computeThreshold()
Definition: HGCalImagingAlgo.cc:628
HGCalImagingAlgo::calculateLocalDensity
double calculateLocalDensity(std::vector< KDNode > &, KDTree &, const unsigned int) const
Definition: HGCalImagingAlgo.cc:262
hgcal_clustering::max_index
size_t max_index(const std::vector< T > &v)
Definition: HGCalClusteringAlgoBase.h:30
CaloCellGeometry.h
HLTEgPhaseIITestSequence_cff.maxNumberOfThickIndices
maxNumberOfThickIndices
Definition: HLTEgPhaseIITestSequence_cff.py:1344
eostools.move
def move(src, dest)
Definition: eostools.py:511
AlignmentPI::calculatePosition
std::pair< double, double > calculatePosition(TVirtualPad *myPad, int boundary)
Definition: AlignmentPayloadInspectorHelper.h:768
KDTreeLinkerAlgo::clear
void clear()
Definition: KDTreeLinkerAlgo.h:122
HGCalImagingAlgo::shareEnergy
void shareEnergy(const std::vector< KDNode > &, const std::vector< unsigned > &, std::vector< std::vector< double >> &)
Definition: HGCalImagingAlgo.cc:531
dqm-mbProfile.log
log
Definition: dqm-mbProfile.py:17
data
char data[epos_bytes_allocation]
Definition: EPOS_Wrapper.h:79
KDTreeLinkerAlgo::build
void build(std::vector< KDTreeNodeInfo< DATA, DIM > > &eltList, const KDTreeBox< DIM > &region)
Definition: KDTreeLinkerAlgo.h:147
funct::pow
Power< A, B >::type pow(const A &a, const B &b)
Definition: Power.h:29
mps_fire.result
result
Definition: mps_fire.py:311
HGCalImagingAlgo::calculatePosition
math::XYZPoint calculatePosition(std::vector< KDNode > &) const
Definition: HGCalImagingAlgo.cc:196
HGCalImagingAlgo.h
dummy
Definition: DummySelector.h:38
dqmiolumiharvest.j
j
Definition: dqmiolumiharvest.py:66
JetChargeProducer_cfi.exp
exp
Definition: JetChargeProducer_cfi.py:6
HLT_FULL_cff.distance
distance
Definition: HLT_FULL_cff.py:7733
HLT_FULL_cff.stoppingTolerance
stoppingTolerance
Definition: HLT_FULL_cff.py:13215
weight
Definition: weight.py:1
MillePedeFileConverter_cfg.e
e
Definition: MillePedeFileConverter_cfg.py:37
HGCalImagingAlgo::populate
void populate(const HGCRecHitCollection &hits) override
Definition: HGCalImagingAlgo.cc:26