mlpack  2.2.5
RASearch< SortPolicy, MetricType, MatType, TreeType > Class Template Reference

The RASearch class: This class provides a generic manner to perform rank-approximate search via random-sampling. More...

Inheritance diagram for RASearch< SortPolicy, MetricType, MatType, TreeType >:

Public Types

typedef TreeType< MetricType, RAQueryStat< SortPolicy >, MatType > Tree
 Convenience typedef. More...

 

Public Member Functions

 RASearch (const MatType &referenceSet, const bool naive=false, const bool singleMode=false, const double tau=5, const double alpha=0.95, const bool sampleAtLeaves=false, const bool firstLeafExact=false, const size_t singleSampleLimit=20, const MetricType metric=MetricType())
 Initialize the RASearch object, passing both a reference dataset (this is the dataset that will be searched). More...

 
 RASearch (MatType &&referenceSet, const bool naive=false, const bool singleMode=false, const double tau=5, const double alpha=0.95, const bool sampleAtLeaves=false, const bool firstLeafExact=false, const size_t singleSampleLimit=20, const MetricType metric=MetricType())
 Initialize the RASearch object, passing both a reference dataset (this is the dataset that will be searched). More...

 
 RASearch (Tree *referenceTree, const bool singleMode=false, const double tau=5, const double alpha=0.95, const bool sampleAtLeaves=false, const bool firstLeafExact=false, const size_t singleSampleLimit=20, const MetricType metric=MetricType())
 Initialize the RASearch object with the given pre-constructed reference tree. More...

 
 RASearch (const bool naive=false, const bool singleMode=false, const double tau=5, const double alpha=0.95, const bool sampleAtLeaves=false, const bool firstLeafExact=false, const size_t singleSampleLimit=20, const MetricType metric=MetricType())
 Create an RASearch object with no reference data. More...

 
 ~RASearch ()
 Delete the RASearch object. More...

 
double Alpha () const
 Get the desired success probability. More...

 
double & Alpha ()
 Modify the desired success probability. More...

 
bool FirstLeafExact () const
 Get whether or not we traverse to the first leaf without approximation. More...

 
bool & FirstLeafExact ()
 Modify whether or not we traverse to the first leaf without approximation. More...

 
bool Naive () const
 Get whether or not naive (brute-force) search is used. More...

 
bool & Naive ()
 Modify whether or not naive (brute-force) search is used. More...

 
const MatType & ReferenceSet () const
 Access the reference set. More...

 
void ResetQueryTree (Tree *queryTree) const
 This function recursively resets the RAQueryStat of the given query tree to set 'bound' to SortPolicy::WorstDistance and 'numSamplesMade' to 0. More...

 
bool SampleAtLeaves () const
 Get whether or not sampling is done at the leaves. More...

 
bool & SampleAtLeaves ()
 Modify whether or not sampling is done at the leaves. More...

 
void Search (const MatType &querySet, const size_t k, arma::Mat< size_t > &neighbors, arma::mat &distances)
 Compute the rank approximate nearest neighbors of each query point in the query set and store the output in the given matrices. More...

 
void Search (Tree *queryTree, const size_t k, arma::Mat< size_t > &neighbors, arma::mat &distances)
 Compute the rank approximate nearest neighbors of each point in the pre-built query tree and store the output in the given matrices. More...

 
void Search (const size_t k, arma::Mat< size_t > &neighbors, arma::mat &distances)
 Compute the rank approximate nearest neighbors of each point in the reference set (that is, the query set is taken to be the reference set), and store the output in the given matrices. More...

 
template
<
typename
Archive
>
void Serialize (Archive &ar, const unsigned int)
 Serialize the object. More...

 
bool SingleMode () const
 Get whether or not single-tree search is used. More...

 
bool & SingleMode ()
 Modify whether or not single-tree search is used. More...

 
size_t SingleSampleLimit () const
 Get the limit on the size of a node that can be approximated. More...

 
size_t & SingleSampleLimit ()
 Modify the limit on the size of a node that can be approximation. More...

 
double Tau () const
 Get the rank-approximation in percentile of the data. More...

 
double & Tau ()
 Modify the rank-approximation in percentile of the data. More...

 
void Train (const MatType &referenceSet)
 "Train" the model on the given reference set. More...

 
void Train (MatType &&referenceSet)
 "Train" the model on the given reference set, taking ownership of the data matrix. More...

 

Detailed Description


template
<
typename
SortPolicy
=
NearestNeighborSort
,
typename
MetricType
=
metric::EuclideanDistance
,
typename
MatType
=
arma::mat
,
template
<
typename
TreeMetricType
,
typename
TreeStatType
,
typename
TreeMatType
>
class
TreeType
=
tree::KDTree
>

class mlpack::neighbor::RASearch< SortPolicy, MetricType, MatType, TreeType >

The RASearch class: This class provides a generic manner to perform rank-approximate search via random-sampling.

If the 'naive' option is chosen, this rank-approximate search will be done by randomly sampling from the whole set. If the 'naive' option is not chosen, the sampling is done in a stratified manner in the tree as mentioned in the algorithms in Figure 2 of the following paper:

{ram2009rank, title={{Rank-Approximate Nearest Neighbor Search: Retaining Meaning and Speed in High Dimensions}}, author={{Ram, P. and Lee, D. and Ouyang, H. and Gray, A. G.}}, booktitle={{Advances of Neural Information Processing Systems}}, year={2009} }

RASearch is currently known to not work with ball trees (#356).

Template Parameters
SortPolicyThe sort policy for distances; see NearestNeighborSort.
MetricTypeThe metric to use for computation.
TreeTypeThe tree type to use.

Definition at line 71 of file ra_search.hpp.

Member Typedef Documentation

◆ Tree

typedef TreeType<MetricType, RAQueryStat<SortPolicy>, MatType> Tree

Convenience typedef.

Definition at line 75 of file ra_search.hpp.

Constructor & Destructor Documentation

◆ RASearch() [1/4]

RASearch ( const MatType &  referenceSet,
const bool  naive = false,
const bool  singleMode = false,
const double  tau = 5,
const double  alpha = 0.95,
const bool  sampleAtLeaves = false,
const bool  firstLeafExact = false,
const size_t  singleSampleLimit = 20,
const MetricType  metric = MetricType() 
)

Initialize the RASearch object, passing both a reference dataset (this is the dataset that will be 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).

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 using the appropriate constructor, or by using the constructor that takes an rvalue reference to the data with std::move().

tau, the rank-approximation parameter, specifies that we are looking for k neighbors with probability alpha of being in the top tau percent of nearest neighbors. So, as an example, if our dataset has 1000 points, and we want 5 nearest neighbors with 95% probability of being in the top 5% of nearest neighbors (or, the top 50 nearest neighbors), we set k = 5, tau = 5, and alpha = 0.95.

The method will fail (and throw a std::invalid_argument exception) if the value of tau is too low: tau must be set such that the number of points in the corresponding percentile of the data is greater than k. Thus, if we choose tau = 0.1 with a dataset of 1000 points and k = 5, then we are attempting to choose 5 nearest neighbors out of the closest 1 point – this is invalid.

Parameters
referenceSetSet of reference points.
naiveIf true, the rank-approximate search will be performed by directly sampling the whole set instead of using the stratified sampling on the tree.
singleModeIf true, single-tree search will be used (as opposed to dual-tree search). This is useful when Search() will be called with few query points.
metricAn optional instance of the MetricType class.
tauThe rank-approximation in percentile of the data. The default value is 5%.
alphaThe desired success probability. The default value is 0.95.
sampleAtLeavesSample at leaves for faster but less accurate computation. This defaults to 'false'.
firstLeafExactTraverse to the first leaf without approximation. This can ensure that the query definitely finds its (near) duplicate if there exists one. This defaults to 'false' for now.
singleSampleLimitThe limit on the largest node that can be approximated by sampling. This defaults to 20.

◆ RASearch() [2/4]

RASearch ( MatType &&  referenceSet,
const bool  naive = false,
const bool  singleMode = false,
const double  tau = 5,
const double  alpha = 0.95,
const bool  sampleAtLeaves = false,
const bool  firstLeafExact = false,
const size_t  singleSampleLimit = 20,
const MetricType  metric = MetricType() 
)

Initialize the RASearch object, passing both a reference dataset (this is the dataset that will be 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).

This method will take ownership of the given reference set, avoiding a copy. If you need to use the reference set for other purposes, too, consider using the constructor that takes a const reference.

tau, the rank-approximation parameter, specifies that we are looking for k neighbors with probability alpha of being in the top tau percent of nearest neighbors. So, as an example, if our dataset has 1000 points, and we want 5 nearest neighbors with 95% probability of being in the top 5% of nearest neighbors (or, the top 50 nearest neighbors), we set k = 5, tau = 5, and alpha = 0.95.

The method will fail (and throw a std::invalid_argument exception) if the value of tau is too low: tau must be set such that the number of points in the corresponding percentile of the data is greater than k. Thus, if we choose tau = 0.1 with a dataset of 1000 points and k = 5, then we are attempting to choose 5 nearest neighbors out of the closest 1 point – this is invalid.

Parameters
referenceSetSet of reference points.
naiveIf true, the rank-approximate search will be performed by directly sampling the whole set instead of using the stratified sampling on the tree.
singleModeIf true, single-tree search will be used (as opposed to dual-tree search). This is useful when Search() will be called with few query points.
metricAn optional instance of the MetricType class.
tauThe rank-approximation in percentile of the data. The default value is 5%.
alphaThe desired success probability. The default value is 0.95.
sampleAtLeavesSample at leaves for faster but less accurate computation. This defaults to 'false'.
firstLeafExactTraverse to the first leaf without approximation. This can ensure that the query definitely finds its (near) duplicate if there exists one. This defaults to 'false' for now.
singleSampleLimitThe limit on the largest node that can be approximated by sampling. This defaults to 20.

◆ RASearch() [3/4]

RASearch ( Tree referenceTree,
const bool  singleMode = false,
const double  tau = 5,
const double  alpha = 0.95,
const bool  sampleAtLeaves = false,
const bool  firstLeafExact = false,
const size_t  singleSampleLimit = 20,
const MetricType  metric = MetricType() 
)

Initialize the RASearch object with the given pre-constructed reference tree.

It is assumed that the points in the tree's dataset correspond to the reference set. Optionally, choose to use single-tree mode. Naive mode is not available as an option for this constructor; instead, to run naive computation, use a different constructor. Additionally, an instantiated distance metric can be given, for cases where the distance metric holds data.

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.

tau, the rank-approximation parameter, specifies that we are looking for k neighbors with probability alpha of being in the top tau percent of nearest neighbors. So, as an example, if our dataset has 1000 points, and we want 5 nearest neighbors with 95% probability of being in the top 5% of nearest neighbors (or, the top 50 nearest neighbors), we set k = 5, tau = 5, and alpha = 0.95.

The method will fail (and throw a std::invalid_argument exception) if the value of tau is too low: tau must be set such that the number of points in the corresponding percentile of the data is greater than k. Thus, if we choose tau = 0.1 with a dataset of 1000 points and k = 5, then we are attempting to choose 5 nearest neighbors out of the closest 1 point – this is invalid.

Note
Tree-building may (at least with BinarySpaceTree) modify the ordering of a matrix, so be aware that the results you get from Search() will correspond to the modified matrix.
Parameters
referenceTreePre-built tree for reference points.
singleModeWhether single-tree computation should be used (as opposed to dual-tree computation).
tauThe rank-approximation in percentile of the data. The default value is 5%.
alphaThe desired success probability. The default value is 0.95.
sampleAtLeavesSample at leaves for faster but less accurate computation. This defaults to 'false'.
firstLeafExactTraverse to the first leaf without approximation. This can ensure that the query definitely finds its (near) duplicate if there exists one. This defaults to 'false' for now.
singleSampleLimitThe limit on the largest node that can be approximated by sampling. This defaults to 20.
metricInstantiated distance metric.

◆ RASearch() [4/4]

RASearch ( const bool  naive = false,
const bool  singleMode = false,
const double  tau = 5,
const double  alpha = 0.95,
const bool  sampleAtLeaves = false,
const bool  firstLeafExact = false,
const size_t  singleSampleLimit = 20,
const MetricType  metric = MetricType() 
)

Create an RASearch object with no reference data.

If Search() is called before a reference set is set with Train(), an exception will be thrown.

Parameters
naiveWhether naive (brute-force) search should be used.
singleModeWhether single-tree computation should be used (as opposed to dual-tree computation).
tauThe rank-approximation in percentile of the data. The default value is 5%.
alphaThe desired success probability. The default value is 0.95.
sampleAtLeavesSample at leaves for faster but less accurate computation. This defaults to 'false'.
firstLeafExactTraverse to the first leaf without approximation. This can ensure that the query definitely finds its (near) duplicate if there exists one. This defaults to 'false' for now.
singleSampleLimitThe limit on the largest node that can be approximated by sampling. This defaults to 20.
metricInstantiated distance metric.

◆ ~RASearch()

~RASearch ( )

Delete the RASearch object.

The tree is the only member we are responsible for deleting. The others will take care of themselves.

Member Function Documentation

◆ Alpha() [1/2]

double Alpha ( ) const
inline

Get the desired success probability.

Definition at line 398 of file ra_search.hpp.

◆ Alpha() [2/2]

double& Alpha ( )
inline

Modify the desired success probability.

Definition at line 400 of file ra_search.hpp.

◆ FirstLeafExact() [1/2]

bool FirstLeafExact ( ) const
inline

Get whether or not we traverse to the first leaf without approximation.

Definition at line 408 of file ra_search.hpp.

◆ FirstLeafExact() [2/2]

bool& FirstLeafExact ( )
inline

Modify whether or not we traverse to the first leaf without approximation.

Definition at line 410 of file ra_search.hpp.

◆ Naive() [1/2]

bool Naive ( ) const
inline

Get whether or not naive (brute-force) search is used.

Definition at line 383 of file ra_search.hpp.

◆ Naive() [2/2]

bool& Naive ( )
inline

Modify whether or not naive (brute-force) search is used.

Definition at line 385 of file ra_search.hpp.

◆ ReferenceSet()

const MatType& ReferenceSet ( ) const
inline

Access the reference set.

Definition at line 380 of file ra_search.hpp.

◆ ResetQueryTree()

void ResetQueryTree ( Tree queryTree) const

This function recursively resets the RAQueryStat of the given query tree to set 'bound' to SortPolicy::WorstDistance and 'numSamplesMade' to 0.

This allows a user to perform multiple searches with the same query tree, possibly with different levels of approximation without requiring to build a new pair of trees for every new (approximate) search.

If Search() is called multiple times with the same query tree without calling ResetQueryTree(), the results may not satisfy the theoretical guarantees provided by the rank-approximate neighbor search algorithm.

Parameters
queryTreeTree whose statistics should be reset.

◆ SampleAtLeaves() [1/2]

bool SampleAtLeaves ( ) const
inline

Get whether or not sampling is done at the leaves.

Definition at line 403 of file ra_search.hpp.

◆ SampleAtLeaves() [2/2]

bool& SampleAtLeaves ( )
inline

Modify whether or not sampling is done at the leaves.

Definition at line 405 of file ra_search.hpp.

◆ Search() [1/3]

void Search ( const MatType &  querySet,
const size_t  k,
arma::Mat< size_t > &  neighbors,
arma::mat &  distances 
)

Compute the rank approximate nearest neighbors of each query point in the query set 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 is small or only contains one point, it can be faster to do single-tree search; single-tree search can be set with the SingleMode() function or in the constructor.

Parameters
querySetSet of query points (can be a single point).
kNumber of neighbors to search for.
neighborsMatrix storing lists of neighbors for each query point.
distancesMatrix storing distances of neighbors for each query point.

◆ Search() [2/3]

void Search ( Tree queryTree,
const size_t  k,
arma::Mat< size_t > &  neighbors,
arma::mat &  distances 
)

Compute the rank approximate nearest neighbors of each point in the pre-built query tree 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 singleMode or naive is enabled, then this method will throw a std::invalid_argument exception; calling this function implies a dual-tree algorithm.

Note
If the tree type you are using modifies the data matrix, be aware that the results returned from this function will be with respect to the modified data matrix.
Parameters
queryTreeTree built on query points.
kNumber of neighbors to search for.
neighborsMatrix storing lists of neighbors for each query point.
distancesMatrix storing distances of neighbors for each query point.

◆ Search() [3/3]

void Search ( const size_t  k,
arma::Mat< size_t > &  neighbors,
arma::mat &  distances 
)

Compute the rank approximate nearest neighbors of each point in the reference set (that is, the query set is taken to be the reference set), 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.

Parameters
kNumber of neighbors to search for.
neighborsMatrix storing lists of neighbors for each point.
distancesMatrix storing distances of neighbors for each query point.

◆ Serialize()

void Serialize ( Archive &  ar,
const unsigned  int 
)

◆ SingleMode() [1/2]

bool SingleMode ( ) const
inline

Get whether or not single-tree search is used.

Definition at line 388 of file ra_search.hpp.

◆ SingleMode() [2/2]

bool& SingleMode ( )
inline

Modify whether or not single-tree search is used.

Definition at line 390 of file ra_search.hpp.

◆ SingleSampleLimit() [1/2]

size_t SingleSampleLimit ( ) const
inline

Get the limit on the size of a node that can be approximated.

Definition at line 413 of file ra_search.hpp.

◆ SingleSampleLimit() [2/2]

size_t& SingleSampleLimit ( )
inline

Modify the limit on the size of a node that can be approximation.

Definition at line 415 of file ra_search.hpp.

◆ Tau() [1/2]

double Tau ( ) const
inline

Get the rank-approximation in percentile of the data.

Definition at line 393 of file ra_search.hpp.

◆ Tau() [2/2]

double& Tau ( )
inline

Modify the rank-approximation in percentile of the data.

Definition at line 395 of file ra_search.hpp.

◆ Train() [1/2]

void Train ( const MatType &  referenceSet)

"Train" the model on the given reference set.

If tree-based search is being used (if Naive() is false), this means rebuilding the reference tree. This particular method will make a copy of the given reference data. To avoid that copy, use the Train() method that takes an rvalue reference with std::move().

Parameters
referenceSetNew reference set to use.

◆ Train() [2/2]

void Train ( MatType &&  referenceSet)

"Train" the model on the given reference set, taking ownership of the data matrix.

If tree-based search is being used (if Naive() is false), this also means rebuilding the reference tree. If you need to keep a copy of the reference data, use the Train() method that takes a const reference to the data.

Parameters
referenceSetNew reference set to use.

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