umappp::Status< Index_, Float_ > Class Template Reference

Status of the UMAP optimization iterations. More...

#include <Status.hpp>

Public Member Functions
int	num_dimensions () const
const Float_ *	embedding () const
void	set_embedding (Float_ *ptr, bool copy=true)
int	epoch () const
int	num_epochs () const
size_t	num_observations () const
void	run (int epoch_limit=0)

Detailed Description

template<typename Index_, typename Float_>
class umappp::Status< Index_, Float_ >

Status of the UMAP optimization iterations.

Template Parameters

Index_	Integer type of the neighbor indices.
Float_	Floating-point type for the distances.

Instances of this class should not be constructed directly, but instead returned by initialize().

Member Function Documentation

embedding()

template<typename Index_ , typename Float_ >

const Float_ * umappp::Status< Index_, Float_ >::embedding ( ) const

inline

Returns

Pointer to an array containing the embeddings after the specified number of epochs. This is a column-major matrix where rows are dimensions (num_dimensions()) and columns are observations (num_observations()).

epoch()

template<typename Index_ , typename Float_ >

int umappp::Status< Index_, Float_ >::epoch ( ) const

inline

Returns

Current epoch.

num_dimensions()

template<typename Index_ , typename Float_ >

int umappp::Status< Index_, Float_ >::num_dimensions ( ) const

inline

Returns

Number of dimensions of the embedding.

num_epochs()

template<typename Index_ , typename Float_ >

int umappp::Status< Index_, Float_ >::num_epochs ( ) const

inline

Returns

Total number of epochs. This is typically determined by the value of Options::max_epochs used in initialize().

num_observations()

template<typename Index_ , typename Float_ >

size_t umappp::Status< Index_, Float_ >::num_observations ( ) const

inline

Returns

The number of observations in the dataset.

run()

template<typename Index_ , typename Float_ >

void umappp::Status< Index_, Float_ >::run ( int  epoch_limit = 0 )

inline

The status of the algorithm and the coordinates in embedding() are updated to the specified number of epochs.

Parameters

epoch_limit

Number of epochs to run to. The actual number of epochs performed is equal to the difference between epoch_limit and the current number of epochs in epoch(). epoch_limit should be not less than epoch() and be no greater than the maximum number of epochs specified in max_epochs(). If zero, defaults to the maximum number of epochs.

set_embedding()

template<typename Index_ , typename Float_ >

void umappp::Status< Index_, Float_ >::set_embedding ( Float_ *  ptr, bool  copy = true  )

inline

Parameters

ptr	Pointer to an array as described in embedding().
copy	Whether the contents of the previous array should be copied into ptr.

By default, the Status object will operate on embeddings in an array specified at its own construction time. This method will change the embedding array for an existing Status object, which can be helpful in some situations, e.g., to clone a Status object and to store its embeddings in a different array than the object.

The contents of the new array in ptr should be the same as the array that it replaces, as run() will continue the iteration from the coordinates inside the array. This is enforced by default when copy = true, but if the supplied ptr already contains a copy, the caller may set copy = false to avoid extra work

---

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

• umappp/Status.hpp