DBSCAN 3D cluster algorithm #139
Conversation
| void clusterizeDBSCAN( const edm::PtrVector<l1t::HGCalCluster> & clustersPtr, | ||
| l1t::HGCalMulticlusterBxCollection & multiclusters); | ||
|
|
||
| void findNeighbor( const edm::PtrVector<l1t::HGCalCluster> & clustersPtrs, |
There was a problem hiding this comment.
If this function is not meant to be used outside the class, put it private
| const l1t::HGCalCluster & cluster, | ||
| std::vector<int> & neighborList); | ||
|
|
||
| bool isNeighbor( const l1t::HGCalCluster & clu1, |
There was a problem hiding this comment.
If this function is not meant to be used outside the class, put it private
| { | ||
| int iclu = 0; | ||
|
|
||
| for(edm::PtrVector<l1t::HGCalCluster>::iterator clu = clustersPtrs.begin(); clu != clustersPtrs.end(); ++clu, ++iclu){ |
There was a problem hiding this comment.
This iterator is const, so better to write it explicitly using const_iterator
| std::vector<std::vector<int>> neighborList; | ||
| int iclu = 0, imclu = 0; | ||
|
|
||
| for(edm::PtrVector<l1t::HGCalCluster>::iterator clu = clustersPtrs.begin(); clu != clustersPtrs.end(); ++clu, ++iclu){ |
There was a problem hiding this comment.
This iterator is const, so better to write it explicitly using const_iterator
| multiclustersTmp.emplace_back( *clu ); | ||
| merged.at(iclu) = true; | ||
|
|
||
| for(unsigned int neighNo = 0; neighNo < neighborList.at(iclu).size(); neighNo++){ |
There was a problem hiding this comment.
A range-based loop would make the code more readable, to avoid having all the neighborList.at(iclu).at(neighNo)
There was a problem hiding this comment.
I spent half a day on this, I don't think range-based loops work here (note that range is not constant). I would be more than happy to try if you have a suggestion.
There was a problem hiding this comment.
Ah yes, right. In that case maybe you can define an integer at the beginning of the loop:
int neighbor = neighborList.at(iclu).at(neighNo);to be used later in the loop.
| @@ -65,7 +65,11 @@ | |||
|
|
|||
| C3d_parValues = cms.PSet( dR_multicluster = cms.double(0.01), # dR in normalized plane used to clusterize C2d | |||
| minPt_multicluster = cms.double(0.5), # minimum pt of the multicluster (GeV) | |||
| calibSF_multicluster = cms.double(1.084) | |||
| calibSF_multicluster = cms.double(1.084), | |||
| type_multicluster = cms.string('DBSCANC3d'), # 'dRC3d' for the cone algorithm | |||
There was a problem hiding this comment.
This algo should not be used as default for the moment since no performance studies have been shown yet.
| calibSF_multicluster = cms.double(1.084) | ||
| calibSF_multicluster = cms.double(1.084), | ||
| type_multicluster = cms.string('DBSCANC3d'), # 'dRC3d' for the cone algorithm | ||
| dist_dbscan_multicluster = cms.double(0.03), |
There was a problem hiding this comment.
If I understand correctly this parameter plays the same role as dR_multicluster. If this is the case one single parameter could be used.
| std::vector<int> neighbors; | ||
|
|
||
| if(!visited.at(iclu)){ | ||
| visited.at(iclu)=true; |
There was a problem hiding this comment.
There are mixed tab and space indents from here down to the end of the function. Please use only spaces.
|
|
||
|
|
||
| void HGCalMulticlusteringImpl::findNeighbor( const edm::PtrVector<l1t::HGCalCluster> & clustersPtrs, | ||
| const l1t::HGCalCluster & cluster, |
There was a problem hiding this comment.
Mixed tab and space indents. Please use only spaces.
| }else { | ||
| edm::LogWarning("ParameterError") << "Unknown Multiclustering type '" << typeMulticluster | ||
| << "'. Using DBSCAN instead.\n"; | ||
| multiclusteringAlgoType_ = DBSCANC3d; |
There was a problem hiding this comment.
This algo should not be used as default for the moment since no performance studies have been shown yet.
|
There are still mixed spaces and tabs |
Implementation of the DBScan algorithm (Ester et al) for linking the 2D clusters.
Required parameters (to be optimized) cluster distance, min number of 2D clusters.
To do: