mlpack::emst::DualTreeBoruvka< MetricType, TreeType > Class Template Reference

Performs the MST calculation using the Dual-Tree Boruvka algorithm, using any type of tree. More...

Collaboration diagram for mlpack::emst::DualTreeBoruvka< MetricType, TreeType >:
Collaboration graph
[legend]

List of all members.

Classes

struct  SortEdgesHelper
 For sorting the edge list after the computation. More...

Public Member Functions

 DualTreeBoruvka (TreeType *tree, const typename TreeType::Mat &dataset, const MetricType metric=MetricType())
 Create the DualTreeBoruvka object with an already initialized tree.
 DualTreeBoruvka (const typename TreeType::Mat &dataset, const bool naive=false, const size_t leafSize=1, const MetricType metric=MetricType())
 Create the tree from the given dataset.
 ~DualTreeBoruvka ()
 Delete the tree, if it was created inside the object.
void ComputeMST (arma::mat &results)
 Iteratively find the nearest neighbor of each component until the MST is complete.

Private Member Functions

void AddAllEdges ()
 Adds all the edges found in one iteration to the list of neighbors.
void AddEdge (const size_t e1, const size_t e2, const double distance)
 Adds a single edge to the edge list.
void Cleanup ()
 The values stored in the tree must be reset on each iteration.
void CleanupHelper (TreeType *tree)
 This function resets the values in the nodes of the tree nearest neighbor distance, and checks for fully connected nodes.
void EmitResults (arma::mat &results)
 Unpermute the edge list and output it to results.

Private Attributes

UnionFind connections
 Connections.
TreeType::Mat & data
 Reference to the data (this is what should be used for accessing data).
TreeType::Mat dataCopy
 Copy of the data (if necessary).
std::vector< EdgePairedges
 Edges.
MetricType metric
 The instantiated metric.
bool naive
 Indicates whether or not O(n^2) naive mode will be used.
arma::vec neighborsDistances
 List of edge distances.
arma::Col< size_t > neighborsInComponent
 List of edge nodes.
arma::Col< size_t > neighborsOutComponent
 List of edge nodes.
std::vector< size_t > oldFromNew
 Permutations of points during tree building.
bool ownTree
 Indicates whether or not we "own" the tree.
struct
mlpack::emst::DualTreeBoruvka::SortEdgesHelper 
SortFun
 For sorting the edge list after the computation.
double totalDist
 Total distance of the tree.
TreeType * tree
 Pointer to the root of the tree.

Detailed Description

template<typename MetricType = metric::EuclideanDistance, typename TreeType = tree::BinarySpaceTree<bound::HRectBound<2>, DTBStat>>
class mlpack::emst::DualTreeBoruvka< MetricType, TreeType >

Performs the MST calculation using the Dual-Tree Boruvka algorithm, using any type of tree.

For more information on the algorithm, see the following citation:

 @inproceedings{
   author = {March, W.B., Ram, P., and Gray, A.G.},
   title = {{Fast Euclidean Minimum Spanning Tree: Algorithm, Analysis,
      Applications.}},
   booktitle = {Proceedings of the 16th ACM SIGKDD International Conference
      on Knowledge Discovery and Data Mining}
   series = {KDD 2010},
   year = {2010}
 }

General usage of this class might be like this:

 extern arma::mat data; // We want to find the MST of this dataset.
 DualTreeBoruvka<> dtb(data); // Create the tree with default options.

 // Find the MST.
 arma::mat mstResults;
 dtb.ComputeMST(mstResults);

More advanced usage of the class can use different types of trees, pass in an already-built tree, or compute the MST using the O(n^2) naive algorithm.

Template Parameters:
MetricType The metric to use. IMPORTANT: this hasn't really been tested with anything other than the L2 metric, so user beware. Note that the tree type needs to compute bounds using the same metric as the type specified here.
TreeType Type of tree to use. Should use DTBStat as a statistic.

Definition at line 91 of file dtb.hpp.


Constructor & Destructor Documentation

template<typename MetricType = metric::EuclideanDistance, typename TreeType = tree::BinarySpaceTree<bound::HRectBound<2>, DTBStat>>
mlpack::emst::DualTreeBoruvka< MetricType, TreeType >::DualTreeBoruvka ( const typename TreeType::Mat &  dataset,
const bool  naive = false,
const size_t  leafSize = 1,
const MetricType  metric = MetricType() 
)

Create the tree from the given dataset.

This copies the dataset to an internal copy, because tree-building modifies the dataset.

Parameters:
data Dataset to build a tree for.
naive Whether the computation should be done in O(n^2) naive mode.
leafSize The leaf size to be used during tree construction.
template<typename MetricType = metric::EuclideanDistance, typename TreeType = tree::BinarySpaceTree<bound::HRectBound<2>, DTBStat>>
mlpack::emst::DualTreeBoruvka< MetricType, TreeType >::DualTreeBoruvka ( TreeType *  tree,
const typename TreeType::Mat &  dataset,
const MetricType  metric = MetricType() 
)

Create the DualTreeBoruvka object with an already initialized tree.

This will not copy the dataset, and can save a little processing power. Naive mode is not available as an option for this constructor; instead, to run naive computation, construct a tree with all the points in one leaf (i.e. leafSize = number of points).

Note:
Because tree-building (at least with BinarySpaceTree) modifies the ordering of a matrix, be sure you pass the modified matrix to this object! In addition, mapping the points of the matrix back to their original indices is not done when this constructor is used.
Parameters:
tree Pre-built tree.
dataset Dataset corresponding to the pre-built tree.
template<typename MetricType = metric::EuclideanDistance, typename TreeType = tree::BinarySpaceTree<bound::HRectBound<2>, DTBStat>>
mlpack::emst::DualTreeBoruvka< MetricType, TreeType >::~DualTreeBoruvka (  ) 

Delete the tree, if it was created inside the object.


Member Function Documentation

template<typename MetricType = metric::EuclideanDistance, typename TreeType = tree::BinarySpaceTree<bound::HRectBound<2>, DTBStat>>
void mlpack::emst::DualTreeBoruvka< MetricType, TreeType >::AddAllEdges (  )  [private]

Adds all the edges found in one iteration to the list of neighbors.

template<typename MetricType = metric::EuclideanDistance, typename TreeType = tree::BinarySpaceTree<bound::HRectBound<2>, DTBStat>>
void mlpack::emst::DualTreeBoruvka< MetricType, TreeType >::AddEdge ( const size_t  e1,
const size_t  e2,
const double  distance 
) [private]

Adds a single edge to the edge list.

template<typename MetricType = metric::EuclideanDistance, typename TreeType = tree::BinarySpaceTree<bound::HRectBound<2>, DTBStat>>
void mlpack::emst::DualTreeBoruvka< MetricType, TreeType >::Cleanup (  )  [private]

The values stored in the tree must be reset on each iteration.

template<typename MetricType = metric::EuclideanDistance, typename TreeType = tree::BinarySpaceTree<bound::HRectBound<2>, DTBStat>>
void mlpack::emst::DualTreeBoruvka< MetricType, TreeType >::CleanupHelper ( TreeType *  tree  )  [private]

This function resets the values in the nodes of the tree nearest neighbor distance, and checks for fully connected nodes.

template<typename MetricType = metric::EuclideanDistance, typename TreeType = tree::BinarySpaceTree<bound::HRectBound<2>, DTBStat>>
void mlpack::emst::DualTreeBoruvka< MetricType, TreeType >::ComputeMST ( arma::mat &  results  ) 

Iteratively find the nearest neighbor of each component until the MST is complete.

The results will be a 3xN matrix (with N equal to the number of edges in the minimum spanning tree). The first row will contain the lesser index of the edge; the second row will contain the greater index of the edge; and the third row will contain the distance between the two edges.

Parameters:
results Matrix which results will be stored in.
template<typename MetricType = metric::EuclideanDistance, typename TreeType = tree::BinarySpaceTree<bound::HRectBound<2>, DTBStat>>
void mlpack::emst::DualTreeBoruvka< MetricType, TreeType >::EmitResults ( arma::mat &  results  )  [private]

Unpermute the edge list and output it to results.


Member Data Documentation

template<typename MetricType = metric::EuclideanDistance, typename TreeType = tree::BinarySpaceTree<bound::HRectBound<2>, DTBStat>>
UnionFind mlpack::emst::DualTreeBoruvka< MetricType, TreeType >::connections [private]

Connections.

Definition at line 111 of file dtb.hpp.

template<typename MetricType = metric::EuclideanDistance, typename TreeType = tree::BinarySpaceTree<bound::HRectBound<2>, DTBStat>>
TreeType::Mat& mlpack::emst::DualTreeBoruvka< MetricType, TreeType >::data [private]

Reference to the data (this is what should be used for accessing data).

Definition at line 97 of file dtb.hpp.

template<typename MetricType = metric::EuclideanDistance, typename TreeType = tree::BinarySpaceTree<bound::HRectBound<2>, DTBStat>>
TreeType::Mat mlpack::emst::DualTreeBoruvka< MetricType, TreeType >::dataCopy [private]

Copy of the data (if necessary).

Definition at line 95 of file dtb.hpp.

template<typename MetricType = metric::EuclideanDistance, typename TreeType = tree::BinarySpaceTree<bound::HRectBound<2>, DTBStat>>
std::vector<EdgePair> mlpack::emst::DualTreeBoruvka< MetricType, TreeType >::edges [private]

Edges.

Definition at line 108 of file dtb.hpp.

template<typename MetricType = metric::EuclideanDistance, typename TreeType = tree::BinarySpaceTree<bound::HRectBound<2>, DTBStat>>
MetricType mlpack::emst::DualTreeBoruvka< MetricType, TreeType >::metric [private]

The instantiated metric.

Definition at line 126 of file dtb.hpp.

template<typename MetricType = metric::EuclideanDistance, typename TreeType = tree::BinarySpaceTree<bound::HRectBound<2>, DTBStat>>
bool mlpack::emst::DualTreeBoruvka< MetricType, TreeType >::naive [private]

Indicates whether or not O(n^2) naive mode will be used.

Definition at line 105 of file dtb.hpp.

template<typename MetricType = metric::EuclideanDistance, typename TreeType = tree::BinarySpaceTree<bound::HRectBound<2>, DTBStat>>
arma::vec mlpack::emst::DualTreeBoruvka< MetricType, TreeType >::neighborsDistances [private]

List of edge distances.

Definition at line 120 of file dtb.hpp.

template<typename MetricType = metric::EuclideanDistance, typename TreeType = tree::BinarySpaceTree<bound::HRectBound<2>, DTBStat>>
arma::Col<size_t> mlpack::emst::DualTreeBoruvka< MetricType, TreeType >::neighborsInComponent [private]

List of edge nodes.

Definition at line 116 of file dtb.hpp.

template<typename MetricType = metric::EuclideanDistance, typename TreeType = tree::BinarySpaceTree<bound::HRectBound<2>, DTBStat>>
arma::Col<size_t> mlpack::emst::DualTreeBoruvka< MetricType, TreeType >::neighborsOutComponent [private]

List of edge nodes.

Definition at line 118 of file dtb.hpp.

template<typename MetricType = metric::EuclideanDistance, typename TreeType = tree::BinarySpaceTree<bound::HRectBound<2>, DTBStat>>
std::vector<size_t> mlpack::emst::DualTreeBoruvka< MetricType, TreeType >::oldFromNew [private]

Permutations of points during tree building.

Definition at line 114 of file dtb.hpp.

template<typename MetricType = metric::EuclideanDistance, typename TreeType = tree::BinarySpaceTree<bound::HRectBound<2>, DTBStat>>
bool mlpack::emst::DualTreeBoruvka< MetricType, TreeType >::ownTree [private]

Indicates whether or not we "own" the tree.

Definition at line 102 of file dtb.hpp.

template<typename MetricType = metric::EuclideanDistance, typename TreeType = tree::BinarySpaceTree<bound::HRectBound<2>, DTBStat>>
struct mlpack::emst::DualTreeBoruvka::SortEdgesHelper mlpack::emst::DualTreeBoruvka< MetricType, TreeType >::SortFun [private]

For sorting the edge list after the computation.

template<typename MetricType = metric::EuclideanDistance, typename TreeType = tree::BinarySpaceTree<bound::HRectBound<2>, DTBStat>>
double mlpack::emst::DualTreeBoruvka< MetricType, TreeType >::totalDist [private]

Total distance of the tree.

Definition at line 123 of file dtb.hpp.

template<typename MetricType = metric::EuclideanDistance, typename TreeType = tree::BinarySpaceTree<bound::HRectBound<2>, DTBStat>>
TreeType* mlpack::emst::DualTreeBoruvka< MetricType, TreeType >::tree [private]

Pointer to the root of the tree.

Definition at line 100 of file dtb.hpp.


The documentation for this class was generated from the following file:

Generated on 29 Sep 2016 for MLPACK by  doxygen 1.6.1