mlpack
2.2.5
|
The NeighborSearch class is a template class for performing distance-based neighbor searches. More...
Public Types | |
typedef TreeType< MetricType, NeighborSearchStat< SortPolicy > , MatType > | Tree |
Convenience typedef. More... | |
Public Member Functions | |
NeighborSearch (const MatType &referenceSet, const NeighborSearchMode mode=DUAL_TREE_MODE, const double epsilon=0, const MetricType metric=MetricType()) | |
Initialize the NeighborSearch object, passing a reference dataset (this is the dataset which is searched). More... | |
NeighborSearch (MatType &&referenceSet, const NeighborSearchMode mode=DUAL_TREE_MODE, const double epsilon=0, const MetricType metric=MetricType()) | |
Initialize the NeighborSearch object, taking ownership of the reference dataset (this is the dataset which is searched). More... | |
NeighborSearch (const Tree &referenceTree, const NeighborSearchMode mode=DUAL_TREE_MODE, const double epsilon=0, const MetricType metric=MetricType()) | |
Initialize the NeighborSearch object with a copy of the given pre-constructed reference tree (this is the tree built on the points that will be searched). More... | |
NeighborSearch (Tree &&referenceTree, const NeighborSearchMode mode=DUAL_TREE_MODE, const double epsilon=0, const MetricType metric=MetricType()) | |
Initialize the NeighborSearch object with the given pre-constructed reference tree (this is the tree built on the points that will be searched). More... | |
NeighborSearch (const NeighborSearchMode mode=DUAL_TREE_MODE, const double epsilon=0, const MetricType metric=MetricType()) | |
Create a NeighborSearch object without any reference data. More... | |
mlpack_deprecated | NeighborSearch (const MatType &referenceSet, const bool naive, const bool singleMode=false, const double epsilon=0, const MetricType metric=MetricType()) |
Initialize the NeighborSearch object, passing a reference dataset (this is the dataset which is searched). More... | |
mlpack_deprecated | NeighborSearch (MatType &&referenceSet, const bool naive, const bool singleMode=false, const double epsilon=0, const MetricType metric=MetricType()) |
Initialize the NeighborSearch object, taking ownership of the reference dataset (this is the dataset which is searched). More... | |
mlpack_deprecated | NeighborSearch (Tree *referenceTree, const bool singleMode, const double epsilon=0, const MetricType metric=MetricType()) |
Initialize the NeighborSearch object with the given pre-constructed reference tree (this is the tree built on the points that will be searched). More... | |
mlpack_deprecated | NeighborSearch (const bool naive, const bool singleMode=false, const double epsilon=0, const MetricType metric=MetricType()) |
Create a NeighborSearch object without any reference data. More... | |
NeighborSearch (const NeighborSearch &other) | |
Copy the given NeighborSearch object (this may consume a lot of memory; be careful!). More... | |
NeighborSearch (NeighborSearch &&other) | |
Take possession of the given NeighborSearch object. More... | |
~NeighborSearch () | |
Delete the NeighborSearch object. More... | |
size_t | BaseCases () const |
Return the total number of base case evaluations performed during the last search. More... | |
double | Epsilon () const |
Access the relative error to be considered in approximate search. More... | |
double & | Epsilon () |
Modify the relative error to be considered in approximate search. More... | |
bool | Greedy () const |
Access whether or not search is done in greedy mode. More... | |
bool & | Greedy () |
Modify whether or not search is done in greedy mode. More... | |
bool | Naive () const |
Access whether or not search is done in naive linear scan mode. More... | |
bool & | Naive () |
Modify whether or not search is done in naive linear scan mode. More... | |
const MatType & | ReferenceSet () const |
Access the reference dataset. More... | |
const Tree & | ReferenceTree () const |
Access the reference tree. More... | |
Tree & | ReferenceTree () |
Modify the reference tree. More... | |
size_t | Scores () const |
Return the number of node combination scores during the last search. More... | |
void | Search (const MatType &querySet, const size_t k, arma::Mat< size_t > &neighbors, arma::mat &distances) |
For each point in the query set, compute the nearest neighbors and store the output in the given matrices. More... | |
mlpack_deprecated void | Search (Tree *queryTree, const size_t k, arma::Mat< size_t > &neighbors, arma::mat &distances, bool sameSet=false) |
Given a pre-built query tree, search for the nearest neighbors of each point in the query tree, storing the output in the given matrices. More... | |
void | Search (Tree &queryTree, const size_t k, arma::Mat< size_t > &neighbors, arma::mat &distances, bool sameSet=false) |
Given a pre-built query tree, search for the nearest neighbors of each point in the query tree, storing the output in the given matrices. More... | |
void | Search (const size_t k, arma::Mat< size_t > &neighbors, arma::mat &distances) |
Search for the nearest neighbors of every point in the reference set. More... | |
template<typename Archive > | |
void | Serialize (Archive &ar, const unsigned int) |
Serialize the NeighborSearch model. More... | |
bool | SingleMode () const |
Access whether or not search is done in single-tree mode. More... | |
bool & | SingleMode () |
Modify whether or not search is done in single-tree mode. More... | |
void | Train (const MatType &referenceSet) |
Set the reference set to a new reference set, and build a tree if necessary. More... | |
void | Train (MatType &&referenceSet) |
Set the reference set to a new reference set, taking ownership of the set, and build a tree if necessary. More... | |
mlpack_deprecated void | Train (Tree *referenceTree) |
Set the reference tree to a new reference tree. More... | |
void | Train (const Tree &referenceTree) |
Set the reference tree as a copy of the given reference tree. More... | |
void | Train (Tree &&referenceTree) |
Set the reference tree to a new reference tree. More... | |
Static Public Member Functions | |
static double | EffectiveError (arma::mat &foundDistances, arma::mat &realDistances) |
Calculate the average relative error (effective error) between the distances calculated and the true distances provided. More... | |
static double | Recall (arma::Mat< size_t > &foundNeighbors, arma::Mat< size_t > &realNeighbors) |
Calculate the recall (% of neighbors found) given the list of found neighbors and the true set of neighbors. More... | |
The NeighborSearch class is a template class for performing distance-based neighbor searches.
It takes a query dataset and a reference dataset (or just a reference dataset) and, for each point in the query dataset, finds the k neighbors in the reference dataset which have the 'best' distance according to a given sorting policy. A constructor is given which takes only a reference dataset, and if that constructor is used, the given reference dataset is also used as the query dataset.
The template parameters SortPolicy and Metric define the sort function used and the metric (distance function) used. More information on those classes can be found in the NearestNeighborSort class and the kernel::ExampleKernel class.
SortPolicy | The sort policy for distances; see NearestNeighborSort. |
MetricType | The metric to use for computation. |
MatType | The type of data matrix. |
TreeType | The tree type to use; must adhere to the TreeType API. |
DualTreeTraversalType | The type of dual tree traversal to use (defaults to the tree's default traverser). |
SingleTreeTraversalType | The type of single tree traversal to use (defaults to the tree's default traverser). |
Definition at line 83 of file neighbor_search.hpp.
typedef TreeType<MetricType, NeighborSearchStat<SortPolicy>, MatType> Tree |
Convenience typedef.
Definition at line 87 of file neighbor_search.hpp.
NeighborSearch | ( | const MatType & | referenceSet, |
const NeighborSearchMode | mode = DUAL_TREE_MODE , |
||
const double | epsilon = 0 , |
||
const MetricType | metric = MetricType() |
||
) |
Initialize the NeighborSearch object, passing a reference dataset (this is the dataset which is searched).
Optionally, perform the computation in a different mode. An initialized distance metric can be given, for cases where the metric has internal data (i.e. the distance::MahalanobisDistance class).
This method will copy the matrices to internal copies, which are rearranged during tree-building. You can avoid this extra copy by pre-constructing the trees and passing them using a different constructor, or by using the construct that takes an rvalue reference to the dataset.
referenceSet | Set of reference points. |
mode | Neighbor search mode. |
epsilon | Relative approximate error (non-negative). |
metric | An optional instance of the MetricType class. |
NeighborSearch | ( | MatType && | referenceSet, |
const NeighborSearchMode | mode = DUAL_TREE_MODE , |
||
const double | epsilon = 0 , |
||
const MetricType | metric = MetricType() |
||
) |
Initialize the NeighborSearch object, taking ownership of the reference dataset (this is the dataset which is searched).
Optionally, perform the computation in a different mode. An initialized distance metric can be given, for cases where the metric has internal data (i.e. the distance::MahalanobisDistance class).
This method will not copy the data matrix, but will take ownership of it, and depending on the type of tree used, may rearrange the points. If you would rather a copy be made, consider using the constructor that takes a const reference to the data instead.
referenceSet | Set of reference points. |
mode | Neighbor search mode. |
epsilon | Relative approximate error (non-negative). |
metric | An optional instance of the MetricType class. |
NeighborSearch | ( | const Tree & | referenceTree, |
const NeighborSearchMode | mode = DUAL_TREE_MODE , |
||
const double | epsilon = 0 , |
||
const MetricType | metric = MetricType() |
||
) |
Initialize the NeighborSearch object with a copy of the given pre-constructed reference tree (this is the tree built on the points that will be searched).
Optionally, choose to use single-tree mode. Naive mode is not available as an option for this constructor. Additionally, an instantiated distance metric can be given, for cases where the distance metric holds data.
This method will copy the given tree. You can avoid this copy by using the construct that takes a rvalue reference to the tree.
referenceTree | Pre-built tree for reference points. |
mode | Neighbor search mode. |
epsilon | Relative approximate error (non-negative). |
metric | Instantiated distance metric. |
NeighborSearch | ( | Tree && | referenceTree, |
const NeighborSearchMode | mode = DUAL_TREE_MODE , |
||
const double | epsilon = 0 , |
||
const MetricType | metric = MetricType() |
||
) |
Initialize the NeighborSearch object with the given pre-constructed reference tree (this is the tree built on the points that will be searched).
Optionally, choose to use single-tree mode. Naive mode is not available as an option for this constructor. Additionally, an instantiated distance metric can be given, for cases where the distance metric holds data.
This method will take ownership of the given tree. There is no copying of the data matrices (because tree-building is not necessary), so this is the constructor to use when copies absolutely must be avoided.
referenceTree | Pre-built tree for reference points. |
mode | Neighbor search mode. |
epsilon | Relative approximate error (non-negative). |
metric | Instantiated distance metric. |
NeighborSearch | ( | const NeighborSearchMode | mode = DUAL_TREE_MODE , |
const double | epsilon = 0 , |
||
const MetricType | metric = MetricType() |
||
) |
Create a NeighborSearch object without any reference data.
If Search() is called before a reference set is set with Train(), an exception will be thrown.
mode | Neighbor search mode. |
epsilon | Relative approximate error (non-negative). |
metric | Instantiated metric. |
mlpack_deprecated NeighborSearch | ( | const MatType & | referenceSet, |
const bool | naive, | ||
const bool | singleMode = false , |
||
const double | epsilon = 0 , |
||
const MetricType | metric = MetricType() |
||
) |
Initialize the NeighborSearch object, passing a reference dataset (this is the dataset which is searched).
Optionally, perform the computation in naive mode or single-tree mode. An initialized distance metric can be given, for cases where the metric has internal data (i.e. the distance::MahalanobisDistance class).
Deprecated. Will be removed in mlpack 3.0.0.
This method will copy the matrices to internal copies, which are rearranged during tree-building. You can avoid this extra copy by pre-constructing the trees and passing them using a different constructor, or by using the construct that takes an rvalue reference to the dataset.
referenceSet | Set of reference points. |
naive | If true, O(n^2) naive search will be used (as opposed to dual-tree search). This overrides singleMode (if it is set to true). |
singleMode | If true, single-tree search will be used (as opposed to dual-tree search). |
epsilon | Relative approximate error (non-negative). |
metric | An optional instance of the MetricType class. |
mlpack_deprecated NeighborSearch | ( | MatType && | referenceSet, |
const bool | naive, | ||
const bool | singleMode = false , |
||
const double | epsilon = 0 , |
||
const MetricType | metric = MetricType() |
||
) |
Initialize the NeighborSearch object, taking ownership of the reference dataset (this is the dataset which is searched).
Optionally, perform the computation in naive mode or single-tree mode. An initialized distance metric can be given, for cases where the metric has internal data (i.e. the distance::MahalanobisDistance class).
Deprecated. Will be removed in mlpack 3.0.0.
This method will not copy the data matrix, but will take ownership of it, and depending on the type of tree used, may rearrange the points. If you would rather a copy be made, consider using the constructor that takes a const reference to the data instead.
referenceSet | Set of reference points. |
naive | If true, O(n^2) naive search will be used (as opposed to dual-tree search). This overrides singleMode (if it is set to true). |
singleMode | If true, single-tree search will be used (as opposed to dual-tree search). |
epsilon | Relative approximate error (non-negative). |
metric | An optional instance of the MetricType class. |
mlpack_deprecated NeighborSearch | ( | Tree * | referenceTree, |
const bool | singleMode, | ||
const double | epsilon = 0 , |
||
const MetricType | metric = MetricType() |
||
) |
Initialize the NeighborSearch object with the given pre-constructed reference tree (this is the tree built on the points that will be searched).
Optionally, choose to use single-tree mode. Naive mode is not available as an option for this constructor. Additionally, an instantiated distance metric can be given, for cases where the distance metric holds data.
Deprecated. Will be removed in mlpack 3.0.0.
This method won't take ownership of the given tree. There is no copying of the data matrices in this constructor (because tree-building is not necessary), so this is the constructor to use when copies absolutely must be avoided.
referenceTree | Pre-built tree for reference points. |
singleMode | Whether single-tree computation should be used (as opposed to dual-tree computation). |
epsilon | Relative approximate error (non-negative). |
metric | Instantiated distance metric. |
mlpack_deprecated NeighborSearch | ( | const bool | naive, |
const bool | singleMode = false , |
||
const double | epsilon = 0 , |
||
const MetricType | metric = MetricType() |
||
) |
Create a NeighborSearch object without any reference data.
If Search() is called before a reference set is set with Train(), an exception will be thrown.
Deprecated. Will be removed in mlpack 3.0.0.
naive | Whether to use naive search. |
singleMode | Whether single-tree computation should be used (as opposed to dual-tree computation). |
epsilon | Relative approximate error (non-negative). |
metric | Instantiated metric. |
NeighborSearch | ( | const NeighborSearch< SortPolicy, MetricType, MatType, TreeType, DualTreeTraversalType, SingleTreeTraversalType > & | other | ) |
Copy the given NeighborSearch object (this may consume a lot of memory; be careful!).
NeighborSearch | ( | NeighborSearch< SortPolicy, MetricType, MatType, TreeType, DualTreeTraversalType, SingleTreeTraversalType > && | other | ) |
Take possession of the given NeighborSearch object.
~NeighborSearch | ( | ) |
Delete the NeighborSearch object.
The tree is the only member we are responsible for deleting. The others will take care of themselves.
|
inline |
Return the total number of base case evaluations performed during the last search.
Definition at line 507 of file neighbor_search.hpp.
|
static |
Calculate the average relative error (effective error) between the distances calculated and the true distances provided.
The input matrices must have the same size.
Cases where the true distance is zero (the same point) or the calculated distance is SortPolicy::WorstDistance() (didn't find enough points) will be ignored.
foundDistances | Matrix storing lists of calculated distances for each query point. |
realDistances | Matrix storing lists of true best distances for each query point. |
|
inline |
Access the relative error to be considered in approximate search.
Definition at line 540 of file neighbor_search.hpp.
|
inline |
Modify the relative error to be considered in approximate search.
Definition at line 542 of file neighbor_search.hpp.
|
inline |
Access whether or not search is done in greedy mode.
Deprecated. Will be replaced in mlpack 3.0.0, by a new method: NeighborSearchMode SearchMode().
Definition at line 533 of file neighbor_search.hpp.
|
inline |
Modify whether or not search is done in greedy mode.
Deprecated. Will be replaced in mlpack 3.0.0, by a new method: NeighborSearchMode& SearchMode().
Definition at line 537 of file neighbor_search.hpp.
|
inline |
Access whether or not search is done in naive linear scan mode.
Deprecated. Will be replaced in mlpack 3.0.0, by a new method: NeighborSearchMode SearchMode().
Definition at line 515 of file neighbor_search.hpp.
|
inline |
Modify whether or not search is done in naive linear scan mode.
Deprecated. Will be replaced in mlpack 3.0.0, by a new method: NeighborSearchMode& SearchMode().
Definition at line 519 of file neighbor_search.hpp.
|
static |
Calculate the recall (% of neighbors found) given the list of found neighbors and the true set of neighbors.
The recall returned will be in the range [0, 1].
foundNeighbors | Matrix storing lists of calculated neighbors for each query point. |
realNeighbors | Matrix storing lists of true best neighbors for each query point. |
|
inline |
Access the reference dataset.
Definition at line 545 of file neighbor_search.hpp.
|
inline |
Access the reference tree.
Definition at line 548 of file neighbor_search.hpp.
|
inline |
Modify the reference tree.
Definition at line 550 of file neighbor_search.hpp.
|
inline |
Return the number of node combination scores during the last search.
Definition at line 510 of file neighbor_search.hpp.
void Search | ( | const MatType & | querySet, |
const size_t | k, | ||
arma::Mat< size_t > & | neighbors, | ||
arma::mat & | distances | ||
) |
For each point in the query set, compute the nearest neighbors and store the output in the given matrices.
The matrices will be set to the size of n columns by k rows, where n is the number of points in the query dataset and k is the number of neighbors being searched for.
If querySet contains only a few query points, the extra cost of building a tree on the points for dual-tree search may not be warranted, and it may be worthwhile to set singleMode = false (either in the constructor or with SingleMode()).
querySet | Set of query points (can be just one point). |
k | Number of neighbors to search for. |
neighbors | Matrix storing lists of neighbors for each query point. |
distances | Matrix storing distances of neighbors for each query point. |
mlpack_deprecated void Search | ( | Tree * | queryTree, |
const size_t | k, | ||
arma::Mat< size_t > & | neighbors, | ||
arma::mat & | distances, | ||
bool | sameSet = false |
||
) |
Given a pre-built query tree, search for the nearest neighbors of each point in the query tree, storing the output in the given matrices.
The matrices will be set to the size of n columns by k rows, where n is the number of points in the query dataset and k is the number of neighbors being searched for.
This method is deprecated and will be removed in mlpack 3.0.0! The Search() method taking a reference to the query tree is preferred.
Note that if you are calling Search() multiple times with a single query tree, you need to reset the bounds in the statistic of each query node, otherwise the result may be wrong! You can do this by calling TreeType::Stat()::Reset() on each node in the query tree.
queryTree | Tree built on query points. |
k | Number of neighbors to search for. |
neighbors | Matrix storing lists of neighbors for each query point. |
distances | Matrix storing distances of neighbors for each query point. |
sameSet | Denotes whether or not the reference and query sets are the same. |
void Search | ( | Tree & | queryTree, |
const size_t | k, | ||
arma::Mat< size_t > & | neighbors, | ||
arma::mat & | distances, | ||
bool | sameSet = false |
||
) |
Given a pre-built query tree, search for the nearest neighbors of each point in the query tree, storing the output in the given matrices.
The matrices will be set to the size of n columns by k rows, where n is the number of points in the query dataset and k is the number of neighbors being searched for.
Note that if you are calling Search() multiple times with a single query tree, you need to reset the bounds in the statistic of each query node, otherwise the result may be wrong! You can do this by calling TreeType::Stat()::Reset() on each node in the query tree.
queryTree | Tree built on query points. |
k | Number of neighbors to search for. |
neighbors | Matrix storing lists of neighbors for each query point. |
distances | Matrix storing distances of neighbors for each query point. |
sameSet | Denotes whether or not the reference and query sets are the same. |
void Search | ( | const size_t | k, |
arma::Mat< size_t > & | neighbors, | ||
arma::mat & | distances | ||
) |
Search for the nearest neighbors of every point in the reference set.
This is basically equivalent to calling any other overload of Search() with the reference set as the query set; so, this lets you do all-k-nearest-neighbors search. The results are stored in the given matrices. The matrices will be set to the size of n columns by k rows, where n is the number of points in the query dataset and k is the number of neighbors being searched for.
k | Number of neighbors to search for. |
neighbors | Matrix storing lists of neighbors for each query point. |
distances | Matrix storing distances of neighbors for each query point. |
void Serialize | ( | Archive & | ar, |
const unsigned | int | ||
) |
Serialize the NeighborSearch model.
|
inline |
Access whether or not search is done in single-tree mode.
Deprecated. Will be replaced in mlpack 3.0.0, by a new method: NeighborSearchMode SearchMode().
Definition at line 524 of file neighbor_search.hpp.
|
inline |
Modify whether or not search is done in single-tree mode.
Deprecated. Will be replaced in mlpack 3.0.0, by a new method: NeighborSearchMode& SearchMode().
Definition at line 528 of file neighbor_search.hpp.
void Train | ( | const MatType & | referenceSet | ) |
Set the reference set to a new reference set, and build a tree if necessary.
This method is called 'Train()' in order to match the rest of the mlpack abstractions, even though calling this "training" is maybe a bit of a stretch.
referenceSet | New set of reference data. |
void Train | ( | MatType && | referenceSet | ) |
Set the reference set to a new reference set, taking ownership of the set, and build a tree if necessary.
This method is called 'Train()' in order to match the rest of the mlpack abstractions, even though calling this "training" is maybe a bit of a stretch.
referenceSet | New set of reference data. |
mlpack_deprecated void Train | ( | Tree * | referenceTree | ) |
Set the reference tree to a new reference tree.
This method is deprecated and will be removed in mlpack 3.0.0! Train() methods taking a reference to the reference tree are preferred.
referenceTree | Pre-built tree for reference points. |
void Train | ( | const Tree & | referenceTree | ) |
Set the reference tree as a copy of the given reference tree.
This method will copy the given tree. You can avoid this copy by using the Train() method that takes a rvalue reference to the tree.
referenceTree | Pre-built tree for reference points. |
void Train | ( | Tree && | referenceTree | ) |
Set the reference tree to a new reference tree.
This method will take ownership of the given tree.
referenceTree | Pre-built tree for reference points. |