lsst::ap::cluster::ClusteringControl Struct Reference

Parameters for the clustering algorithm and its internals. More...

#include <ClusteringControl.h>

Public Member Functions
	ClusteringControl ()
	~ClusteringControl ()
lsst::afw::geom::Angle const	getEpsilon () const
lsst::afw::geom::Angle const	getLeafExtentThreshold () const
void	validate () const

Public Attributes
double	epsilonArcsec
	"Clustering distance (arcsec) to use when generating clusters with\n" "the OPTICS algorithm. If a source S has at least minNeighbors\n" "other sources within an angular separation of epsilonArcsec, then\n" "it is always assigned to a cluster.\n" ; More...
int	minNeighbors
	"The minimum cardinality of the epsilonArcsec-neighborhood of a source\n" "S for S to be considered a core-source by the OPTICS algorithm. Core\n" "sources are always assigned to a cluster, whereas sources with\n" "epsilonArcsec-neighborhoods containing less than minNeighbors sources\n" "may or may not be. If such a source is assigned to a cluster, it is\n" "called a border source. Otherwise, it is called a noise source.\n" "\n" "This parameter is essentially a lower bound on the number of times\n" "an astrophysical object must be detected before its detections are\n" "clustered and turned into a catalog entry and can be tuned to avoid\n" "generating too many spurious entries. To ensure that every source is\n" "assigned to a cluster, set the value to zero. However, setting the\n" "value to a non-negligeable fraction of the number of times the sky is\n" "covered by the data-set in question will typically result in better\n" "clusters. There is currently no way to adjust the value to account\n" "for data-sets with non uniform coverage.\n" ; More...
int	pointsPerLeaf
	"A performance tuning parameter for the k-d tree used internally by\n" "the OPTICS implementation. The height of the tree is picked such\n" "that no leaf will contain more than pointsPerLeaf sources. A value\n" "in the tens of sources is generally a good pick.\n" ; More...
double	leafExtentThresholdArcsec
	"A performance tuning parameter for the k-d tree used internally by\n" "the OPTICS implementation. Nodes that have a maximum extent below\n" "this threshold value in each dimension are not subdivided. The value\n" "should be of the same order as epsilonArcsec - nodes much smaller\n" "than this are useless in the sense that the sources belonging to\n" "such a node become increasingly likely to all lie in the\n" "neighborhood of a query point. Note also that the k-d tree\n" "implementation does not store bounding boxes for nodes, meaning\n" "that an entire node cannot be determined to satisfy a range\n" "query without traversal of its children/contents. This saves on tree\n" "size, and makes sense for the target use-case because query regions\n" "are typically very small.\n" ; More...

Detailed Description

Parameters for the clustering algorithm and its internals.

Definition at line 40 of file ClusteringControl.h.

Constructor & Destructor Documentation

lsst::ap::cluster::ClusteringControl::ClusteringControl

(

)

Definition at line 38 of file ClusteringControl.cc.

                                     :
    epsilonArcsec(0.75),
    minNeighbors(2),
    pointsPerLeaf(32),
    leafExtentThresholdArcsec(2.0)
{
    validate();
}

lsst::ap::cluster::ClusteringControl::~ClusteringControl

(

)

Definition at line 47 of file ClusteringControl.cc.

{ }

Member Function Documentation

lsst::afw::geom::Angle const lsst::ap::cluster::ClusteringControl::getEpsilon ( ) const

inline

Definition at line 88 of file ClusteringControl.h.

                                              {
        return epsilonArcsec * lsst::afw::geom::arcseconds;
    }

lsst::afw::geom::Angle const lsst::ap::cluster::ClusteringControl::getLeafExtentThreshold ( ) const

inline

Definition at line 92 of file ClusteringControl.h.

                                                          {
        return leafExtentThresholdArcsec * lsst::afw::geom::arcseconds;
    }

void lsst::ap::cluster::ClusteringControl::validate

(

)

const

Definition at line 49 of file ClusteringControl.cc.

                                       {
    if (epsilonArcsec < 0.0 || epsilonArcsec > 36000.0) {
        throw LSST_EXCEPT(InvalidParameterError,
                          "epsilonArcsec must lie in the range [0, 36000]");
    }
    if (minNeighbors < 0) {
        throw LSST_EXCEPT(InvalidParameterError,
                          "minNeighbors must be non-negative");

    }
    if (pointsPerLeaf <= 0) {
        throw LSST_EXCEPT(InvalidParameterError,
                          "pointsPerLeaf must be positive");
    }
}

Member Data Documentation

double lsst::ap::cluster::ClusteringControl::epsilonArcsec

"Clustering distance (arcsec) to use when generating clusters with\n" "the OPTICS algorithm. If a source S has at least minNeighbors\n" "other sources within an angular separation of epsilonArcsec, then\n" "it is always assigned to a cluster.\n" ;

Definition at line 48 of file ClusteringControl.h.

double lsst::ap::cluster::ClusteringControl::leafExtentThresholdArcsec

"A performance tuning parameter for the k-d tree used internally by\n" "the OPTICS implementation. Nodes that have a maximum extent below\n" "this threshold value in each dimension are not subdivided. The value\n" "should be of the same order as epsilonArcsec - nodes much smaller\n" "than this are useless in the sense that the sources belonging to\n" "such a node become increasingly likely to all lie in the\n" "neighborhood of a query point. Note also that the k-d tree\n" "implementation does not store bounding boxes for nodes, meaning\n" "that an entire node cannot be determined to satisfy a range\n" "query without traversal of its children/contents. This saves on tree\n" "size, and makes sense for the target use-case because query regions\n" "are typically very small.\n" ;

Definition at line 86 of file ClusteringControl.h.

int lsst::ap::cluster::ClusteringControl::minNeighbors

"The minimum cardinality of the epsilonArcsec-neighborhood of a source\n" "S for S to be considered a core-source by the OPTICS algorithm. Core\n" "sources are always assigned to a cluster, whereas sources with\n" "epsilonArcsec-neighborhoods containing less than minNeighbors sources\n" "may or may not be. If such a source is assigned to a cluster, it is\n" "called a border source. Otherwise, it is called a noise source.\n" "\n" "This parameter is essentially a lower bound on the number of times\n" "an astrophysical object must be detected before its detections are\n" "clustered and turned into a catalog entry and can be tuned to avoid\n" "generating too many spurious entries. To ensure that every source is\n" "assigned to a cluster, set the value to zero. However, setting the\n" "value to a non-negligeable fraction of the number of times the sky is\n" "covered by the data-set in question will typically result in better\n" "clusters. There is currently no way to adjust the value to account\n" "for data-sets with non uniform coverage.\n" ;

Definition at line 66 of file ClusteringControl.h.

int lsst::ap::cluster::ClusteringControl::pointsPerLeaf

"A performance tuning parameter for the k-d tree used internally by\n" "the OPTICS implementation. The height of the tree is picked such\n" "that no leaf will contain more than pointsPerLeaf sources. A value\n" "in the tens of sources is generally a good pick.\n" ;

Definition at line 72 of file ClusteringControl.h.

---

The documentation for this struct was generated from the following files:

• /home/lsstsw/stack/Linux64/ap/10.0+224/include/lsst/ap/cluster/ClusteringControl.h
• /home/lsstsw/stack/Linux64/ap/10.0+224/src/cluster/ClusteringControl.cc