Back to home page

Project CMSSW displayed by LXR
 
 

    

File indexing completed on 2023-03-17 11:18:04 

0001 // Author: Felice Pantaleo, Marco Rovere - felice.pantaleo@cern.ch, marco.rovere@cern.ch
0002 // Date: 11/2018
0003 #include <algorithm>
0004 #include <set>
0005 #include <vector>
0006 
0007 #include "FWCore/MessageLogger/interface/MessageLogger.h"
0008 #include "FWCore/Utilities/interface/Exception.h"
0009 #include "PatternRecognitionbyCA.h"
0010 
0011 #include "TrackstersPCA.h"
0012 #include "Geometry/CaloGeometry/interface/CaloGeometry.h"
0013 #include "Geometry/Records/interface/CaloGeometryRecord.h"
0014 #include "FWCore/Framework/interface/EventSetup.h"
0015 
0016 using namespace ticl;
0017 
0018 template <typename TILES>
0019 PatternRecognitionbyCA<TILES>::PatternRecognitionbyCA(const edm::ParameterSet &conf, edm::ConsumesCollector iC)
0020     : PatternRecognitionAlgoBaseT<TILES>(conf, iC),
0021       caloGeomToken_(iC.esConsumes<CaloGeometry, CaloGeometryRecord>()),
0022       theGraph_(std::make_unique<HGCGraphT<TILES>>()),
0023       oneTracksterPerTrackSeed_(conf.getParameter<bool>("oneTracksterPerTrackSeed")),
0024       promoteEmptyRegionToTrackster_(conf.getParameter<bool>("promoteEmptyRegionToTrackster")),
0025       out_in_dfs_(conf.getParameter<bool>("out_in_dfs")),
0026       max_out_in_hops_(conf.getParameter<int>("max_out_in_hops")),
0027       min_cos_theta_(conf.getParameter<double>("min_cos_theta")),
0028       min_cos_pointing_(conf.getParameter<double>("min_cos_pointing")),
0029       root_doublet_max_distance_from_seed_squared_(
0030           conf.getParameter<double>("root_doublet_max_distance_from_seed_squared")),
0031       etaLimitIncreaseWindow_(conf.getParameter<double>("etaLimitIncreaseWindow")),
0032       skip_layers_(conf.getParameter<int>("skip_layers")),
0033       max_missing_layers_in_trackster_(conf.getParameter<int>("max_missing_layers_in_trackster")),
0034       check_missing_layers_(max_missing_layers_in_trackster_ < 100),
0035       shower_start_max_layer_(conf.getParameter<int>("shower_start_max_layer")),
0036       min_layers_per_trackster_(conf.getParameter<int>("min_layers_per_trackster")),
0037       filter_on_categories_(conf.getParameter<std::vector<int>>("filter_on_categories")),
0038       pid_threshold_(conf.getParameter<double>("pid_threshold")),
0039       energy_em_over_total_threshold_(conf.getParameter<double>("energy_em_over_total_threshold")),
0040       max_longitudinal_sigmaPCA_(conf.getParameter<double>("max_longitudinal_sigmaPCA")),
0041       min_clusters_per_ntuplet_(min_layers_per_trackster_),
0042       max_delta_time_(conf.getParameter<double>("max_delta_time")),
0043       eidInputName_(conf.getParameter<std::string>("eid_input_name")),
0044       eidOutputNameEnergy_(conf.getParameter<std::string>("eid_output_name_energy")),
0045       eidOutputNameId_(conf.getParameter<std::string>("eid_output_name_id")),
0046       eidMinClusterEnergy_(conf.getParameter<double>("eid_min_cluster_energy")),
0047       eidNLayers_(conf.getParameter<int>("eid_n_layers")),
0048       eidNClusters_(conf.getParameter<int>("eid_n_clusters")),
0049       siblings_maxRSquared_(conf.getParameter<std::vector<double>>("siblings_maxRSquared")){};
0050 
0051 template <typename TILES>
0052 PatternRecognitionbyCA<TILES>::~PatternRecognitionbyCA(){};
0053 
0054 template <typename TILES>
0055 void PatternRecognitionbyCA<TILES>::makeTracksters(
0056     const typename PatternRecognitionAlgoBaseT<TILES>::Inputs &input,
0057     std::vector<Trackster> &result,
0058     std::unordered_map<int, std::vector<int>> &seedToTracksterAssociation) {
0059   // Protect from events with no seeding regions
0060   if (input.regions.empty())
0061     return;
0062 
0063   edm::EventSetup const &es = input.es;
0064   const CaloGeometry &geom = es.getData(caloGeomToken_);
0065   rhtools_.setGeometry(geom);
0066 
0067   theGraph_->setVerbosity(PatternRecognitionAlgoBaseT<TILES>::algo_verbosity_);
0068   theGraph_->clear();
0069   if (PatternRecognitionAlgoBaseT<TILES>::algo_verbosity_ > VerbosityLevel::None) {
0070     LogDebug("HGCPatternRecoByCA") << "Making Tracksters with CA" << std::endl;
0071   }
0072 
0073   constexpr auto isHFnose = std::is_same<TILES, TICLLayerTilesHFNose>::value;
0074   constexpr int nEtaBin = TILES::constants_type_t::nEtaBins;
0075   constexpr int nPhiBin = TILES::constants_type_t::nPhiBins;
0076 
0077   bool isRegionalIter = (input.regions[0].index != -1);
0078   std::vector<HGCDoublet::HGCntuplet> foundNtuplets;
0079   std::vector<int> seedIndices;
0080   std::vector<uint8_t> layer_cluster_usage(input.layerClusters.size(), 0);
0081   theGraph_->makeAndConnectDoublets(input.tiles,
0082                                     input.regions,
0083                                     nEtaBin,
0084                                     nPhiBin,
0085                                     input.layerClusters,
0086                                     input.mask,
0087                                     input.layerClustersTime,
0088                                     1,
0089                                     1,
0090                                     min_cos_theta_,
0091                                     min_cos_pointing_,
0092                                     root_doublet_max_distance_from_seed_squared_,
0093                                     etaLimitIncreaseWindow_,
0094                                     skip_layers_,
0095                                     rhtools_.lastLayer(isHFnose),
0096                                     max_delta_time_,
0097                                     rhtools_.lastLayerEE(isHFnose),
0098                                     rhtools_.lastLayerFH(),
0099                                     siblings_maxRSquared_);
0100 
0101   theGraph_->findNtuplets(foundNtuplets, seedIndices, min_clusters_per_ntuplet_, out_in_dfs_, max_out_in_hops_);
0102   //#ifdef FP_DEBUG
0103   const auto &doublets = theGraph_->getAllDoublets();
0104   int tracksterId = -1;
0105 
0106   // container for holding tracksters before selection
0107   std::vector<Trackster> tmpTracksters;
0108   tmpTracksters.reserve(foundNtuplets.size());
0109 
0110   for (auto const &ntuplet : foundNtuplets) {
0111     tracksterId++;
0112 
0113     std::set<unsigned int> effective_cluster_idx;
0114 
0115     for (auto const &doublet : ntuplet) {
0116       auto innerCluster = doublets[doublet].innerClusterId();
0117       auto outerCluster = doublets[doublet].outerClusterId();
0118 
0119       effective_cluster_idx.insert(innerCluster);
0120       effective_cluster_idx.insert(outerCluster);
0121 
0122       if (PatternRecognitionAlgoBaseT<TILES>::algo_verbosity_ > VerbosityLevel::Advanced) {
0123         LogDebug("HGCPatternRecoByCA") << " New doublet " << doublet << " for trackster: " << result.size()
0124                                        << " InnerCl " << innerCluster << " " << input.layerClusters[innerCluster].x()
0125                                        << " " << input.layerClusters[innerCluster].y() << " "
0126                                        << input.layerClusters[innerCluster].z() << " OuterCl " << outerCluster << " "
0127                                        << input.layerClusters[outerCluster].x() << " "
0128                                        << input.layerClusters[outerCluster].y() << " "
0129                                        << input.layerClusters[outerCluster].z() << " " << tracksterId << std::endl;
0130       }
0131     }
0132     unsigned showerMinLayerId = 99999;
0133     std::vector<unsigned int> uniqueLayerIds;
0134     uniqueLayerIds.reserve(effective_cluster_idx.size());
0135     std::vector<std::pair<unsigned int, unsigned int>> lcIdAndLayer;
0136     lcIdAndLayer.reserve(effective_cluster_idx.size());
0137     for (auto const i : effective_cluster_idx) {
0138       auto const &haf = input.layerClusters[i].hitsAndFractions();
0139       auto layerId = rhtools_.getLayerWithOffset(haf[0].first);
0140       showerMinLayerId = std::min(layerId, showerMinLayerId);
0141       uniqueLayerIds.push_back(layerId);
0142       lcIdAndLayer.emplace_back(i, layerId);
0143     }
0144     std::sort(uniqueLayerIds.begin(), uniqueLayerIds.end());
0145     uniqueLayerIds.erase(std::unique(uniqueLayerIds.begin(), uniqueLayerIds.end()), uniqueLayerIds.end());
0146     unsigned int numberOfLayersInTrackster = uniqueLayerIds.size();
0147     if (check_missing_layers_) {
0148       int numberOfMissingLayers = 0;
0149       unsigned int j = showerMinLayerId;
0150       unsigned int indexInVec = 0;
0151       for (const auto &layer : uniqueLayerIds) {
0152         if (layer != j) {
0153           numberOfMissingLayers++;
0154           j++;
0155           if (numberOfMissingLayers > max_missing_layers_in_trackster_) {
0156             numberOfLayersInTrackster = indexInVec;
0157             for (auto &llpair : lcIdAndLayer) {
0158               if (llpair.second >= layer) {
0159                 effective_cluster_idx.erase(llpair.first);
0160               }
0161             }
0162             break;
0163           }
0164         }
0165         indexInVec++;
0166         j++;
0167       }
0168     }
0169     if ((numberOfLayersInTrackster >= min_layers_per_trackster_) and (showerMinLayerId <= shower_start_max_layer_)) {
0170       // Put back indices, in the form of a Trackster, into the results vector
0171       Trackster tmp;
0172       tmp.vertices().reserve(effective_cluster_idx.size());
0173       tmp.vertex_multiplicity().resize(effective_cluster_idx.size(), 1);
0174       //regions and seedIndices can have different size
0175       //if a seeding region does not lead to any trackster
0176       tmp.setSeed(input.regions[0].collectionID, seedIndices[tracksterId]);
0177 
0178       std::copy(std::begin(effective_cluster_idx), std::end(effective_cluster_idx), std::back_inserter(tmp.vertices()));
0179       tmpTracksters.push_back(tmp);
0180     }
0181   }
0182   ticl::assignPCAtoTracksters(tmpTracksters,
0183                               input.layerClusters,
0184                               input.layerClustersTime,
0185                               rhtools_.getPositionLayer(rhtools_.lastLayerEE(isHFnose), isHFnose).z());
0186 
0187   // run energy regression and ID
0188   energyRegressionAndID(input.layerClusters, input.tfSession, tmpTracksters);
0189   // Filter results based on PID criteria or EM/Total energy ratio.
0190   // We want to **keep** tracksters whose cumulative
0191   // probability summed up over the selected categories
0192   // is greater than the chosen threshold. Therefore
0193   // the filtering function should **discard** all
0194   // tracksters **below** the threshold.
0195   auto filter_on_pids = [&](Trackster &t) -> bool {
0196     auto cumulative_prob = 0.;
0197     for (auto index : filter_on_categories_) {
0198       cumulative_prob += t.id_probabilities(index);
0199     }
0200     return (cumulative_prob <= pid_threshold_) &&
0201            (t.raw_em_energy() < energy_em_over_total_threshold_ * t.raw_energy());
0202   };
0203 
0204   std::vector<unsigned int> selectedTrackstersIds;
0205   for (unsigned i = 0; i < tmpTracksters.size(); ++i) {
0206     if (!filter_on_pids(tmpTracksters[i]) and tmpTracksters[i].sigmasPCA()[0] < max_longitudinal_sigmaPCA_) {
0207       selectedTrackstersIds.push_back(i);
0208     }
0209   }
0210 
0211   result.reserve(selectedTrackstersIds.size());
0212 
0213   for (unsigned i = 0; i < selectedTrackstersIds.size(); ++i) {
0214     const auto &t = tmpTracksters[selectedTrackstersIds[i]];
0215     for (auto const lcId : t.vertices()) {
0216       layer_cluster_usage[lcId]++;
0217       if (PatternRecognitionAlgoBaseT<TILES>::algo_verbosity_ > VerbosityLevel::Basic)
0218         LogDebug("HGCPatternRecoByCA") << "LayerID: " << lcId << " count: " << (int)layer_cluster_usage[lcId]
0219                                        << std::endl;
0220     }
0221     if (isRegionalIter) {
0222       seedToTracksterAssociation[t.seedIndex()].push_back(i);
0223     }
0224     result.push_back(t);
0225   }
0226 
0227   for (auto &trackster : result) {
0228     assert(trackster.vertices().size() <= trackster.vertex_multiplicity().size());
0229     for (size_t i = 0; i < trackster.vertices().size(); ++i) {
0230       trackster.vertex_multiplicity()[i] = layer_cluster_usage[trackster.vertices(i)];
0231       if (PatternRecognitionAlgoBaseT<TILES>::algo_verbosity_ > VerbosityLevel::Basic)
0232         LogDebug("HGCPatternRecoByCA") << "LayerID: " << trackster.vertices(i)
0233                                        << " count: " << (int)trackster.vertex_multiplicity(i) << std::endl;
0234     }
0235   }
0236   // Now decide if the tracksters from the track-based iterations have to be merged
0237   if (oneTracksterPerTrackSeed_) {
0238     std::vector<Trackster> tmp;
0239     mergeTrackstersTRK(result, input.layerClusters, tmp, seedToTracksterAssociation);
0240     tmp.swap(result);
0241   }
0242 
0243   ticl::assignPCAtoTracksters(result,
0244                               input.layerClusters,
0245                               input.layerClustersTime,
0246                               rhtools_.getPositionLayer(rhtools_.lastLayerEE(isHFnose), isHFnose).z());
0247 
0248   // run energy regression and ID
0249   energyRegressionAndID(input.layerClusters, input.tfSession, result);
0250 
0251   // now adding dummy tracksters from seeds not connected to any shower in the result collection
0252   // these are marked as charged hadrons with probability 1.
0253   if (promoteEmptyRegionToTrackster_) {
0254     emptyTrackstersFromSeedsTRK(result, seedToTracksterAssociation, input.regions[0].collectionID);
0255   }
0256 
0257   if (PatternRecognitionAlgoBaseT<TILES>::algo_verbosity_ > VerbosityLevel::Advanced) {
0258     for (auto &trackster : result) {
0259       LogDebug("HGCPatternRecoByCA") << "Trackster characteristics: " << std::endl;
0260       LogDebug("HGCPatternRecoByCA") << "Size: " << trackster.vertices().size() << std::endl;
0261       auto counter = 0;
0262       for (auto const &p : trackster.id_probabilities()) {
0263         LogDebug("HGCPatternRecoByCA") << counter++ << ": " << p << std::endl;
0264       }
0265     }
0266   }
0267   theGraph_->clear();
0268 }
0269 
0270 template <typename TILES>
0271 void PatternRecognitionbyCA<TILES>::mergeTrackstersTRK(
0272     const std::vector<Trackster> &input,
0273     const std::vector<reco::CaloCluster> &layerClusters,
0274     std::vector<Trackster> &output,
0275     std::unordered_map<int, std::vector<int>> &seedToTracksterAssociation) const {
0276   output.reserve(input.size());
0277   for (auto &thisSeed : seedToTracksterAssociation) {
0278     auto &tracksters = thisSeed.second;
0279     if (!tracksters.empty()) {
0280       auto numberOfTrackstersInSeed = tracksters.size();
0281       output.emplace_back(input[tracksters[0]]);
0282       auto &outTrackster = output.back();
0283       tracksters[0] = output.size() - 1;
0284       auto updated_size = outTrackster.vertices().size();
0285       for (unsigned int j = 1; j < numberOfTrackstersInSeed; ++j) {
0286         auto &thisTrackster = input[tracksters[j]];
0287         updated_size += thisTrackster.vertices().size();
0288         if (PatternRecognitionAlgoBaseT<TILES>::algo_verbosity_ > VerbosityLevel::Basic) {
0289           LogDebug("HGCPatternRecoByCA") << "Updated size: " << updated_size << std::endl;
0290         }
0291         outTrackster.vertices().reserve(updated_size);
0292         outTrackster.vertex_multiplicity().reserve(updated_size);
0293         std::copy(std::begin(thisTrackster.vertices()),
0294                   std::end(thisTrackster.vertices()),
0295                   std::back_inserter(outTrackster.vertices()));
0296         std::copy(std::begin(thisTrackster.vertex_multiplicity()),
0297                   std::end(thisTrackster.vertex_multiplicity()),
0298                   std::back_inserter(outTrackster.vertex_multiplicity()));
0299       }
0300       tracksters.resize(1);
0301 
0302       // Find duplicate LCs
0303       auto &orig_vtx = outTrackster.vertices();
0304       auto vtx_sorted{orig_vtx};
0305       std::sort(std::begin(vtx_sorted), std::end(vtx_sorted));
0306       for (unsigned int iLC = 1; iLC < vtx_sorted.size(); ++iLC) {
0307         if (vtx_sorted[iLC] == vtx_sorted[iLC - 1]) {
0308           // Clean up duplicate LCs
0309           const auto lcIdx = vtx_sorted[iLC];
0310           const auto firstEl = std::find(orig_vtx.begin(), orig_vtx.end(), lcIdx);
0311           const auto firstPos = std::distance(std::begin(orig_vtx), firstEl);
0312           auto iDup = std::find(std::next(firstEl), orig_vtx.end(), lcIdx);
0313           while (iDup != orig_vtx.end()) {
0314             orig_vtx.erase(iDup);
0315             outTrackster.vertex_multiplicity().erase(outTrackster.vertex_multiplicity().begin() +
0316                                                      std::distance(std::begin(orig_vtx), iDup));
0317             outTrackster.vertex_multiplicity()[firstPos] -= 1;
0318             iDup = std::find(std::next(firstEl), orig_vtx.end(), lcIdx);
0319           };
0320         }
0321       }
0322     }
0323   }
0324   output.shrink_to_fit();
0325 }
0326 
0327 template <typename TILES>
0328 void PatternRecognitionbyCA<TILES>::emptyTrackstersFromSeedsTRK(
0329     std::vector<Trackster> &tracksters,
0330     std::unordered_map<int, std::vector<int>> &seedToTracksterAssociation,
0331     const edm::ProductID &collectionID) const {
0332   for (auto &thisSeed : seedToTracksterAssociation) {
0333     if (thisSeed.second.empty()) {
0334       Trackster t;
0335       t.setRegressedEnergy(0.f);
0336       t.zeroProbabilities();
0337       t.setIdProbability(ticl::Trackster::ParticleType::charged_hadron, 1.f);
0338       t.setSeed(collectionID, thisSeed.first);
0339       tracksters.emplace_back(t);
0340       thisSeed.second.emplace_back(tracksters.size() - 1);
0341     }
0342   }
0343 }
0344 
0345 template <typename TILES>
0346 void PatternRecognitionbyCA<TILES>::energyRegressionAndID(const std::vector<reco::CaloCluster> &layerClusters,
0347                                                           const tensorflow::Session *eidSession,
0348                                                           std::vector<Trackster> &tracksters) {
0349   // Energy regression and particle identification strategy:
0350   //
0351   // 1. Set default values for regressed energy and particle id for each trackster.
0352   // 2. Store indices of tracksters whose total sum of cluster energies is above the
0353   //    eidMinClusterEnergy_ (GeV) threshold. Inference is not applied for soft tracksters.
0354   // 3. When no trackster passes the selection, return.
0355   // 4. Create input and output tensors. The batch dimension is determined by the number of
0356   //    selected tracksters.
0357   // 5. Fill input tensors with layer cluster features. Per layer, clusters are ordered descending
0358   //    by energy. Given that tensor data is contiguous in memory, we can use pointer arithmetic to
0359   //    fill values, even with batching.
0360   // 6. Zero-fill features for empty clusters in each layer.
0361   // 7. Batched inference.
0362   // 8. Assign the regressed energy and id probabilities to each trackster.
0363   //
0364   // Indices used throughout this method:
0365   // i -> batch element / trackster
0366   // j -> layer
0367   // k -> cluster
0368   // l -> feature
0369 
0370   // set default values per trackster, determine if the cluster energy threshold is passed,
0371   // and store indices of hard tracksters
0372   std::vector<int> tracksterIndices;
0373   for (int i = 0; i < (int)tracksters.size(); i++) {
0374     // calculate the cluster energy sum (2)
0375     // note: after the loop, sumClusterEnergy might be just above the threshold which is enough to
0376     // decide whether to run inference for the trackster or not
0377     float sumClusterEnergy = 0.;
0378     for (const unsigned int &vertex : tracksters[i].vertices()) {
0379       sumClusterEnergy += (float)layerClusters[vertex].energy();
0380       // there might be many clusters, so try to stop early
0381       if (sumClusterEnergy >= eidMinClusterEnergy_) {
0382         // set default values (1)
0383         tracksters[i].setRegressedEnergy(0.f);
0384         tracksters[i].zeroProbabilities();
0385         tracksterIndices.push_back(i);
0386         break;
0387       }
0388     }
0389   }
0390 
0391   // do nothing when no trackster passes the selection (3)
0392   int batchSize = (int)tracksterIndices.size();
0393   if (batchSize == 0) {
0394     return;
0395   }
0396 
0397   // create input and output tensors (4)
0398   tensorflow::TensorShape shape({batchSize, eidNLayers_, eidNClusters_, eidNFeatures_});
0399   tensorflow::Tensor input(tensorflow::DT_FLOAT, shape);
0400   tensorflow::NamedTensorList inputList = {{eidInputName_, input}};
0401 
0402   std::vector<tensorflow::Tensor> outputs;
0403   std::vector<std::string> outputNames;
0404   if (!eidOutputNameEnergy_.empty()) {
0405     outputNames.push_back(eidOutputNameEnergy_);
0406   }
0407   if (!eidOutputNameId_.empty()) {
0408     outputNames.push_back(eidOutputNameId_);
0409   }
0410 
0411   // fill input tensor (5)
0412   for (int i = 0; i < batchSize; i++) {
0413     const Trackster &trackster = tracksters[tracksterIndices[i]];
0414 
0415     // per layer, we only consider the first eidNClusters_ clusters in terms of energy, so in order
0416     // to avoid creating large / nested structures to do the sorting for an unknown number of total
0417     // clusters, create a sorted list of layer cluster indices to keep track of the filled clusters
0418     std::vector<int> clusterIndices(trackster.vertices().size());
0419     for (int k = 0; k < (int)trackster.vertices().size(); k++) {
0420       clusterIndices[k] = k;
0421     }
0422     sort(clusterIndices.begin(), clusterIndices.end(), [&layerClusters, &trackster](const int &a, const int &b) {
0423       return layerClusters[trackster.vertices(a)].energy() > layerClusters[trackster.vertices(b)].energy();
0424     });
0425 
0426     // keep track of the number of seen clusters per layer
0427     std::vector<int> seenClusters(eidNLayers_);
0428 
0429     // loop through clusters by descending energy
0430     for (const int &k : clusterIndices) {
0431       // get features per layer and cluster and store the values directly in the input tensor
0432       const reco::CaloCluster &cluster = layerClusters[trackster.vertices(k)];
0433       int j = rhtools_.getLayerWithOffset(cluster.hitsAndFractions()[0].first) - 1;
0434       if (j < eidNLayers_ && seenClusters[j] < eidNClusters_) {
0435         // get the pointer to the first feature value for the current batch, layer and cluster
0436         float *features = &input.tensor<float, 4>()(i, j, seenClusters[j], 0);
0437 
0438         // fill features
0439         *(features++) = float(cluster.energy() / float(trackster.vertex_multiplicity(k)));
0440         *(features++) = float(std::abs(cluster.eta()));
0441         *(features) = float(cluster.phi());
0442 
0443         // increment seen clusters
0444         seenClusters[j]++;
0445       }
0446     }
0447 
0448     // zero-fill features of empty clusters in each layer (6)
0449     for (int j = 0; j < eidNLayers_; j++) {
0450       for (int k = seenClusters[j]; k < eidNClusters_; k++) {
0451         float *features = &input.tensor<float, 4>()(i, j, k, 0);
0452         for (int l = 0; l < eidNFeatures_; l++) {
0453           *(features++) = 0.f;
0454         }
0455       }
0456     }
0457   }
0458 
0459   // run the inference (7)
0460   tensorflow::run(eidSession, inputList, outputNames, &outputs);
0461 
0462   // store regressed energy per trackster (8)
0463   if (!eidOutputNameEnergy_.empty()) {
0464     // get the pointer to the energy tensor, dimension is batch x 1
0465     float *energy = outputs[0].flat<float>().data();
0466 
0467     for (const int &i : tracksterIndices) {
0468       tracksters[i].setRegressedEnergy(*(energy++));
0469     }
0470   }
0471 
0472   // store id probabilities per trackster (8)
0473   if (!eidOutputNameId_.empty()) {
0474     // get the pointer to the id probability tensor, dimension is batch x id_probabilities.size()
0475     int probsIdx = eidOutputNameEnergy_.empty() ? 0 : 1;
0476     float *probs = outputs[probsIdx].flat<float>().data();
0477 
0478     for (const int &i : tracksterIndices) {
0479       tracksters[i].setProbabilities(probs);
0480       probs += tracksters[i].id_probabilities().size();
0481     }
0482   }
0483 }
0484 
0485 template <typename TILES>
0486 void PatternRecognitionbyCA<TILES>::fillPSetDescription(edm::ParameterSetDescription &iDesc) {
0487   iDesc.add<int>("algo_verbosity", 0);
0488   iDesc.add<bool>("oneTracksterPerTrackSeed", false);
0489   iDesc.add<bool>("promoteEmptyRegionToTrackster", false);
0490   iDesc.add<bool>("out_in_dfs", true);
0491   iDesc.add<int>("max_out_in_hops", 10);
0492   iDesc.add<double>("min_cos_theta", 0.915);
0493   iDesc.add<double>("min_cos_pointing", -1.);
0494   iDesc.add<double>("root_doublet_max_distance_from_seed_squared", 9999);
0495   iDesc.add<double>("etaLimitIncreaseWindow", 2.1);
0496   iDesc.add<int>("skip_layers", 0);
0497   iDesc.add<int>("max_missing_layers_in_trackster", 9999);
0498   iDesc.add<int>("shower_start_max_layer", 9999)->setComment("make default such that no filtering is applied");
0499   iDesc.add<int>("min_layers_per_trackster", 10);
0500   iDesc.add<std::vector<int>>("filter_on_categories", {0});
0501   iDesc.add<double>("pid_threshold", 0.)->setComment("make default such that no filtering is applied");
0502   iDesc.add<double>("energy_em_over_total_threshold", -1.)
0503       ->setComment("make default such that no filtering is applied");
0504   iDesc.add<double>("max_longitudinal_sigmaPCA", 9999);
0505   iDesc.add<double>("max_delta_time", 3.)->setComment("nsigma");
0506   iDesc.add<std::string>("eid_input_name", "input");
0507   iDesc.add<std::string>("eid_output_name_energy", "output/regressed_energy");
0508   iDesc.add<std::string>("eid_output_name_id", "output/id_probabilities");
0509   iDesc.add<double>("eid_min_cluster_energy", 1.);
0510   iDesc.add<int>("eid_n_layers", 50);
0511   iDesc.add<int>("eid_n_clusters", 10);
0512   iDesc.add<std::vector<double>>("siblings_maxRSquared", {6e-4, 6e-4, 6e-4});
0513 }
0514 
0515 template class ticl::PatternRecognitionbyCA<TICLLayerTiles>;
0516 template class ticl::PatternRecognitionbyCA<TICLLayerTilesHFNose>;
[ Source navigation ] [ Diff markup ] [ Identifier search ] [ general search ]

This page was automatically generated by the 2.2.1 LXR engine.

The LXR team
Valid CSS 2.1! Valid HTML 4.01!