LMNN< MetricType, OptimizerType > Class Template Reference

An implementation of Large Margin nearest neighbor metric learning technique. More...

Public Member Functions
	LMNN (const arma::mat &dataset, const arma::Row< size_t > &labels, const size_t k, const MetricType metric=MetricType())
	Initialize the LMNN object, passing a dataset (distance metric is learned using this dataset) and labels. More...
const arma::mat &	Dataset () const
	Get the dataset reference. More...
const size_t &	K () const
	Access the value of k. More...
size_t	K ()
	Modify the value of k. More...
const arma::Row< size_t > &	Labels () const
	Get the labels reference. More...
template<typename... CallbackTypes>
void	LearnDistance (arma::mat &outputMatrix, CallbackTypes &&... callbacks)
	Perform Large Margin Nearest Neighbors metric learning. More...
const OptimizerType &	Optimizer () const
	Get the optimizer. More...
OptimizerType &	Optimizer ()
const size_t &	Range () const
	Access the range value. More...
size_t &	Range ()
	Modify the range value. More...
const double &	Regularization () const
	Access the regularization value. More...
double &	Regularization ()
	Modify the regularization value. More...

Detailed Description

template

<

typename

MetricType

=

metric::SquaredEuclideanDistance

,

typename

OptimizerType

=

ens::AMSGrad

>

class mlpack::lmnn::LMNN< MetricType, OptimizerType >

An implementation of Large Margin nearest neighbor metric learning technique.

The method seeks to improve clustering & classification algorithms on a dataset by transforming the dataset representation in a more convenient form for them. It introduces the concept of target neighbors and impostors, focusing on the idea that the distance between impostors and the perimeters established by target neighbors should be large and that between target neighbors and data point should be small. It requires the knowledge of target neighbors beforehand. Moreover, target neighbors once initialized remain same.

For more details, see the following published paper:

@ARTICLE{weinberger09distance,
  author = {Weinberger, K.Q. and Saul, L.K.},
  title = {{Distance metric learning for large margin nearest neighbor
      classification}},
  journal = {The Journal of Machine Learning Research},
  year = {2009},
  volume = {10},
  pages = {207--244},
  publisher = {MIT Press}
}

Template Parameters

MetricType	The type of metric to use for computation.
OptimizerType	Optimizer to use for developing distance.

Definition at line 55 of file lmnn.hpp.

Constructor & Destructor Documentation

LMNN()

LMNN	(	const arma::mat &	dataset,
		const arma::Row< size_t > &	labels,
		const size_t	k,
		const MetricType	metric = MetricType()
	)

Initialize the LMNN object, passing a dataset (distance metric is learned using this dataset) and labels.

Initialization will copy both dataset and labels matrices to internal copies.

Parameters

dataset	Input dataset.
labels	Input dataset labels.
k	Number of targets to consider.
metric	Type of metric used for computation.

Member Function Documentation

Dataset()

const arma::mat& Dataset ( ) const

inline

Get the dataset reference.

Definition at line 90 of file lmnn.hpp.

K() [1/2]

const size_t& K ( ) const

inline

Access the value of k.

Definition at line 106 of file lmnn.hpp.

K() [2/2]

size_t K ( )

inline

Modify the value of k.

Definition at line 108 of file lmnn.hpp.

Labels()

const arma::Row<size_t>& Labels ( ) const

inline

Get the labels reference.

Definition at line 93 of file lmnn.hpp.

LearnDistance()

void LearnDistance	(	arma::mat &	outputMatrix,
		CallbackTypes &&...	callbacks
	)

Perform Large Margin Nearest Neighbors metric learning.

The output distance matrix is written into the passed reference. If the LearnDistance() is called with an outputMatrix with correct dimensions, then that matrix will be used as the starting point for optimization.

Template Parameters

CallbackTypes

Types of Callback functions.

Parameters

outputMatrix	Covariance matrix of Mahalanobis distance.
callbacks	Callback function for ensmallen optimizer OptimizerType. See https://www.ensmallen.org/docs.html#callback-documentation.

Optimizer() [1/2]

const OptimizerType& Optimizer ( ) const

inline

Get the optimizer.

Definition at line 111 of file lmnn.hpp.

Optimizer() [2/2]

OptimizerType& Optimizer ( )

inline

Definition at line 112 of file lmnn.hpp.

Range() [1/2]

const size_t& Range ( ) const

inline

Access the range value.

Definition at line 101 of file lmnn.hpp.

Range() [2/2]

size_t& Range ( )

inline

Modify the range value.

Definition at line 103 of file lmnn.hpp.

Regularization() [1/2]

const double& Regularization ( ) const

inline

Access the regularization value.

Definition at line 96 of file lmnn.hpp.

Regularization() [2/2]

double& Regularization ( )

inline

Modify the regularization value.

Definition at line 98 of file lmnn.hpp.

---

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

• /home/ryan/src/mlpack.org/_src/mlpack-git/src/mlpack/methods/lmnn/lmnn.hpp