Implements data access to data cached entirely in memory, concrete file type support is delegated to derived classes. More...

#include <data_accessor_splitting_mem.hpp>

Inheritance diagram for FST::Data_Accessor_Splitting_Mem< DATATYPE, IDXTYPE, INTERVALCONTAINER >:

Public Types
typedef Data_Accessor_Splitting< DATATYPE, IDXTYPE, INTERVALCONTAINER >	DAS
	the immediate base class, aliased for brevity in derived classes
typedef std::shared_ptr< Data_Scaler< DATATYPE > >	PScaler
	shared pointer to the scaler applied to loaded data
typedef DAS::PPattern	PPattern
	inherited from Data_Accessor: raw pointer to one read-only feature vector
typedef DAS::PSplitters	PSplitters
Public Types inherited from FST::Data_Accessor_Splitting< DATATYPE, IDXTYPE, INTERVALCONTAINER >
typedef std::shared_ptr< Data_Splitter< INTERVALCONTAINER, IDXTYPE > >	PSplitter
	shared pointer to one Data_Splitter (one nested splitting level)
typedef std::shared_ptr< std::vector< PSplitter > >	PSplitters
	shared pointer to the ordered list of splitters, one per nested splitting depth (depth 0 = outermost)
typedef const DATATYPE *	PPattern
Public Types inherited from FST::Data_Accessor< DATATYPE, IDXTYPE >
typedef const DATATYPE *	PPattern
	one data sample: a raw pointer to its (read-only) feature vector, getNoOfFeatures() entries long

Public Member Functions
	Data_Accessor_Splitting_Mem (const std::string _filename, const PSplitters _dsp, const PScaler _dsc)
virtual void	initialize ()
	to be called after object creation (for initialization of data file access etc.)
void	set_scaler_fit_train (const bool b)
	Switches on per-split scaling: the scaler learns its transform on the current depth-0 split's TRAIN part only and applies it to ALL data, refreshed on every depth-0 split change.
bool	get_scaler_fit_train () const
	see set_scaler_fit_train()
void	scale_sample_inplace (DATATYPE *sample) const
	Applies the learned per-feature transform to `sample`, a caller-owned vector of getNoOfFeatures() values, in place.
virtual bool	getFirstSplit () override
	data access iteration (to support, e.g., loops in cross-validation)
virtual bool	getNextSplit () override
	data access iteration (to support, e.g., loops in cross-validation)
virtual bool	getFirstBlock (const DataPart ofwhat, PPattern &firstpattern, IDXTYPE &patterns, const unsigned int loopdepth=0) override
	returns pointer to first consecutive block of data of requested DataPart type in the current split (access iteration)
virtual bool	getNextBlock (const DataPart ofwhat, PPattern &firstpattern, IDXTYPE &patterns, const unsigned int loopdepth=0) override
	returns pointer to next consecutive block of data of requested DataPart type in the current split (access iteration)
Data_Accessor_Splitting_Mem *	clone () const override
Data_Accessor_Splitting_Mem *	sharing_clone () const override=0
Data_Accessor_Splitting_Mem *	stateless_clone () const override
virtual std::ostream &	print (std::ostream &os) const override
	writes the splitter stack's print() output (or "Data_Accessor_Splitting()" if the stack is empty) to `os`
Public Member Functions inherited from FST::Data_Accessor_Splitting< DATATYPE, IDXTYPE, INTERVALCONTAINER >
	Data_Accessor_Splitting (const PSplitters _dsp)
	Constructs an (uninitialized) accessor bound to the splitter stack `_dsp`; concrete subclasses must still call initialize() once dimensionality/class sizes are known.
virtual unsigned int	getNoOfClasses () const override
	returns number of classes
virtual unsigned int	getNoOfFeatures () const override
	returns data dimensionality
virtual IDXTYPE	getClassSize (const unsigned int c) const override
	returns size (number of samples in) of class c
virtual IDXTYPE	getClassSizeSum () const override
	returns summed size (number of samples in) of all classes, i.e., no. of all patterns in data
virtual void	setClass (const int c) override
	sets active class -> from now on only data from class c will be considered
virtual int	getClass () const override
	returns active class
void	setSplittingDepth (const unsigned int depth)
	Selects the nested splitting level (0 = outermost/depth-0 splitter, 1 = the next-nested level, etc.) that subsequent getFirstSplit()/getNextSplit()/getFirstBlock()/getNextBlock() calls operate on.
unsigned int	getSplittingDepth () const
	the nested splitting level last set via setSplittingDepth()
virtual unsigned int	getNoOfSplits () const override
	data access iteration (to support, e.g., loops in cross-validation)
virtual unsigned int	getSplitIndex () const override
	data access iteration (to support, e.g., loops in cross-validation)
virtual IDXTYPE	getNoOfBlocks (const DataPart ofwhat) const override
virtual IDXTYPE	getBlockIndex (const unsigned int loopdepth=0) const override
	returns index of the current consecutive block of data of requested DataPart type in the current split (access iteration)
virtual IDXTYPE	getNoOfPatterns (const DataPart ofwhat) const override
	returns number of patterns in all consecutive blocks of data of requested DataPart type in the current split (access iteration)
virtual void	substitute (const DataPart source, const DataPart target) override
	enables change of meaning of DataPart types, for use in specialized data access scenarios like in bias predicting wrappers
virtual void	resubstitute () override
	resets standard DataPart types' meaning
Public Member Functions inherited from FST::Data_Accessor< DATATYPE, IDXTYPE >
Block_Range	blocks (const DataPart ofwhat)
	RAII block range over the current split - prefer over the deprecated getFirstBlock()/getNextBlock() cursor.
Public Member Functions inherited from FST::Clonable
bool	is_clone () const
bool	is_sharing () const
	true if this instance is a sharing_clone() result (referencing, not owning, the source's structures)
bool	is_sharing_clone () const
	true if this instance is BOTH a clone and sharing – i.e. specifically a sharing_clone() result, not a clone()/stateless_clone() one

Protected Member Functions
	Data_Accessor_Splitting_Mem (const Data_Accessor_Splitting_Mem &damt, int)
	weak (referencing) copy-constructor to be used in sharing_clone()
virtual void	initial_data_read ()=0
	fills the already-allocated data array with the file's samples; called once by initialize() after the array is sized Must write exactly datasize values, laid out class by class, pattern by pattern, feature by feature (class c starts at _class_offset[c]), matching the counts the file_get*() queries report.
virtual void	initial_file_prepare ()=0
	opens/prepares the underlying file for reading; called once by initialize() before initial_data_read() After it returns, the three file_get*() queries below must answer correctly.
virtual unsigned int	file_getNoOfClasses () const =0
	number of classes as reported by the underlying file (valid only after initial_file_prepare())
virtual unsigned int	file_getNoOfFeatures () const =0
	data dimensionality as reported by the underlying file (valid only after initial_file_prepare())
virtual IDXTYPE	file_getClassSize (unsigned int cls) const =0
	number of samples of class `cls` as reported by the underlying file (valid only after initial_file_prepare())
void	initial_data_scaling ()
void	per_split_data_scaling ()
	re-learns the scaler transform on the current depth-0 TRAIN part and re-scales the whole working buffer from the pristine data (scaler_fit_train mode only)
Protected Member Functions inherited from FST::Data_Accessor_Splitting< DATATYPE, IDXTYPE, INTERVALCONTAINER >
	Data_Accessor_Splitting (const Data_Accessor_Splitting &da)
	"Strong" (1:1, deep) copy constructor: stateless_clone()s every splitter and re-assign()s it to this instance's own (deep-copied via splits) interval lists.
void	initialize (const unsigned int _features, const CLASSSIZES &_classes)
DataPart	mappedDataPart (const DataPart ofwhat) const
	looks up `ofwhat's` current remapping target (mappedTRAIN/mappedTEST/mappedTRAINTEST/mappedALL); NONE maps to itself
virtual bool	getFirstBlock (const DataPart ofwhat, DATAINTERVAL &tmp, const unsigned int loopdepth=0)
	returns Data_Interval record representing the first consecutive block of data of requested DataPart type in the current split (access iteration)
virtual bool	getNextBlock (const DataPart ofwhat, DATAINTERVAL &tmp, const unsigned int loopdepth=0)
	returns Data_Interval record representing the next consecutive block of data of requested DataPart type in the current split (access iteration)
bool	is_initialized () const
void	assert_splits (const int splitting_check=-1) const
	DEBUG-only: asserts the internal structures (dsp, splits, enum_split, enum_block, tt_phase, classes) are non-empty, consistently sized, and every splits entry has its train/test/reduced interval lists allocated; a no-op in release builds.
Protected Member Functions inherited from FST::Clonable
void	set_cloned (const bool cloned=true)
	marks/unmarks this instance as a clone; call from clone()/stateless_clone() overrides (Clonable_Impl does this automatically)
void	set_sharing (const bool sharing=true)
	marks/unmarks this instance as sharing (see is_sharing())
void	set_sharing_cloned (const bool cloned=true, const bool sharing=true)
	sets both flags at once; call from sharing_clone() overrides (Clonable_Impl does this automatically)

Protected Attributes
PScaler	dsc
std::vector< PScaler >	feature_scalers
std::string	filename
IDXTYPE	datasize
	total number of DATATYPE values held in data/data_raw (sum over classes of class size * features)
std::shared_ptr< DATATYPE >	data
std::shared_ptr< DATATYPE >	data_raw
	pristine (unscaled) data as loaded; allocated in scaler_fit_train mode only; shared across clones, never mutated after load
bool	scaler_fit_train
	per-split scaling mode switch (see set_scaler_fit_train())
std::vector< IDXTYPE >	_class_offset
	class start indexes (offsets) in whatever random-access data representation
Protected Attributes inherited from FST::Data_Accessor_Splitting< DATATYPE, IDXTYPE, INTERVALCONTAINER >
CLASSSIZES	classes
	pattern count per class, as passed to initialize()
unsigned int	features
	data dimensionality, as passed to initialize()
DataPart	mappedTRAIN
	current remapping target of TRAIN (see substitute()/resubstitute()); defaults to TRAIN
DataPart	mappedTEST
	current remapping target of TEST (see substitute()/resubstitute()); defaults to TEST
DataPart	mappedTRAINTEST
	current remapping target of TRAINTEST (see substitute()/resubstitute()); defaults to TRAINTEST
DataPart	mappedALL
	current remapping target of ALL (see substitute()/resubstitute()); defaults to ALL
PSplitters	dsp
	the splitter stack this accessor was constructed with, one Data_Splitter per nested splitting depth
std::vector< std::vector< DataSplit > >	splits
	one set of splitters per each splitting depth and class
std::vector< IDXTYPE >	enum_split
	current split.. 0~none
std::vector< std::vector< IDXTYPE > >	enum_block
	current block loop.. 0~none
std::vector< std::vector< DataPart > >	tt_phase
	in current block loop.. for DataPart==TRAINTEST indicates which half is being delivered: TRAIN, then TEST once the train blocks run out, NONE when the loop is finished
unsigned int	splitting_depth
	the nested splitting level the getSplit()/getBlock() calls currently refer to; set via setSplittingDepth()
int	active_class
	denotes from which class the getTrain getValidate getTest methods return patterns, to be set using setClass()
bool	_initialize_called
	set true by initialize(); read by is_initialized()

Additional Inherited Members
Protected Types inherited from FST::Data_Accessor_Splitting< DATATYPE, IDXTYPE, INTERVALCONTAINER >
typedef std::vector< unsigned int >	CLASSSIZES
	per-class pattern counts, indexed by class index; the format initialize() takes
typedef const Data_Interval< IDXTYPE > *	DATAINTERVAL
typedef std::shared_ptr< INTERVALCONTAINER >	PIntervaller

Detailed Description

template<typename DATATYPE, typename IDXTYPE, class INTERVALCONTAINER>
class FST::Data_Accessor_Splitting_Mem< DATATYPE, IDXTYPE, INTERVALCONTAINER >

Implements data access to data cached entirely in memory, concrete file type support is delegated to derived classes.

The whole data set is read once into one contiguous DATATYPE array. Classes are stored one after another (class c starts at _class_offset[c]), each class holds its patterns consecutively, and each pattern holds its getNoOfFeatures() values consecutively. A block handed out by getFirstBlock()/getNextBlock() is therefore just a pointer into that array plus a pattern count; the array stays alive for the accessor's lifetime and is never handed over to the caller.

A derived class supplies the file format: it implements initial_file_prepare() (open and read the header), initial_data_read() (fill the array) and the three file_get*() queries that report dimensionality and class sizes. initialize() drives all of them, in that order, and must be called exactly once before any data access.

Member Typedef Documentation

◆ PSplitters

template<typename DATATYPE, typename IDXTYPE, class INTERVALCONTAINER>

typedef DAS::PSplitters FST::Data_Accessor_Splitting_Mem< DATATYPE, IDXTYPE, INTERVALCONTAINER >::PSplitters

inherited: shared pointer to the per-depth splitter stack

Constructor & Destructor Documentation

◆ Data_Accessor_Splitting_Mem()

template<typename DATATYPE, typename IDXTYPE, class INTERVALCONTAINER>

FST::Data_Accessor_Splitting_Mem< DATATYPE, IDXTYPE, INTERVALCONTAINER >::Data_Accessor_Splitting_Mem	(	const std::string	_filename,
		const PSplitters	_dsp,
		const PScaler	_dsc )

inline

Constructs the accessor for _filename (opened by a derived class's initial_file_prepare()), the splitter stack _dsp and the scaler _dsc; actual file reading happens in initialize().

Member Function Documentation

◆ clone()

template<typename DATATYPE, typename IDXTYPE, class INTERVALCONTAINER>

Data_Accessor_Splitting_Mem * FST::Data_Accessor_Splitting_Mem< DATATYPE, IDXTYPE, INTERVALCONTAINER >::clone ( ) const

inlineoverridevirtual

create 1:1 independent clone of the current object. The caller owns the returned object. Families that cannot support a given clone kind implement it to throw fst_error; Clonable_Impl's CLONE_INHERIT slot defers to such a body.

Implements FST::Data_Accessor< DATATYPE, IDXTYPE >.

◆ file_getClassSize()

template<typename DATATYPE, typename IDXTYPE, class INTERVALCONTAINER>

virtual IDXTYPE FST::Data_Accessor_Splitting_Mem< DATATYPE, IDXTYPE, INTERVALCONTAINER >::file_getClassSize ( unsigned int cls ) const

protectedpure virtual

number of samples of class cls as reported by the underlying file (valid only after initial_file_prepare())

Implemented in FST::Data_Accessor_Splitting_MemARFF< DATATYPE, IDXTYPE, INTERVALCONTAINER >, FST::Data_Accessor_Splitting_MemCSV< DATATYPE, IDXTYPE, INTERVALCONTAINER >, FST::Data_Accessor_Splitting_MemTRN< DATATYPE, IDXTYPE, INTERVALCONTAINER >, and FST::Data_Accessor_Universe< DATATYPE, IDXTYPE, INTERVALCONTAINER >.

◆ file_getNoOfClasses()

template<typename DATATYPE, typename IDXTYPE, class INTERVALCONTAINER>

virtual unsigned int FST::Data_Accessor_Splitting_Mem< DATATYPE, IDXTYPE, INTERVALCONTAINER >::file_getNoOfClasses ( ) const

protectedpure virtual

number of classes as reported by the underlying file (valid only after initial_file_prepare())

Implemented in FST::Data_Accessor_Splitting_MemARFF< DATATYPE, IDXTYPE, INTERVALCONTAINER >, FST::Data_Accessor_Splitting_MemCSV< DATATYPE, IDXTYPE, INTERVALCONTAINER >, FST::Data_Accessor_Splitting_MemTRN< DATATYPE, IDXTYPE, INTERVALCONTAINER >, and FST::Data_Accessor_Universe< DATATYPE, IDXTYPE, INTERVALCONTAINER >.

◆ file_getNoOfFeatures()

template<typename DATATYPE, typename IDXTYPE, class INTERVALCONTAINER>

virtual unsigned int FST::Data_Accessor_Splitting_Mem< DATATYPE, IDXTYPE, INTERVALCONTAINER >::file_getNoOfFeatures ( ) const

protectedpure virtual

data dimensionality as reported by the underlying file (valid only after initial_file_prepare())

Implemented in FST::Data_Accessor_Splitting_MemARFF< DATATYPE, IDXTYPE, INTERVALCONTAINER >, FST::Data_Accessor_Splitting_MemCSV< DATATYPE, IDXTYPE, INTERVALCONTAINER >, FST::Data_Accessor_Splitting_MemTRN< DATATYPE, IDXTYPE, INTERVALCONTAINER >, and FST::Data_Accessor_Universe< DATATYPE, IDXTYPE, INTERVALCONTAINER >.

◆ getFirstBlock()

template<typename DATATYPE, typename IDXTYPE, class INTERVALCONTAINER>

bool FST::Data_Accessor_Splitting_Mem< DATATYPE, IDXTYPE, INTERVALCONTAINER >::getFirstBlock	(	const DataPart	ofwhat,
		PPattern &	firstpattern,
		IDXTYPE &	patterns,
		const unsigned int	loopdepth = 0 )

overridevirtual

returns pointer to first consecutive block of data of requested DataPart type in the current split (access iteration)

Implements FST::Data_Accessor_Splitting< DATATYPE, IDXTYPE, INTERVALCONTAINER >.

◆ getFirstSplit()

template<typename DATATYPE, typename IDXTYPE, class INTERVALCONTAINER>

bool FST::Data_Accessor_Splitting_Mem< DATATYPE, IDXTYPE, INTERVALCONTAINER >::getFirstSplit ( )

overridevirtual

data access iteration (to support, e.g., loops in cross-validation)

Reimplemented from FST::Data_Accessor_Splitting< DATATYPE, IDXTYPE, INTERVALCONTAINER >.

Reimplemented in FST::Data_Accessor_Universe< DATATYPE, IDXTYPE, INTERVALCONTAINER >.

◆ getNextBlock()

template<typename DATATYPE, typename IDXTYPE, class INTERVALCONTAINER>

bool FST::Data_Accessor_Splitting_Mem< DATATYPE, IDXTYPE, INTERVALCONTAINER >::getNextBlock	(	const DataPart	ofwhat,
		PPattern &	firstpattern,
		IDXTYPE &	patterns,
		const unsigned int	loopdepth = 0 )

overridevirtual

returns pointer to next consecutive block of data of requested DataPart type in the current split (access iteration)

Implements FST::Data_Accessor_Splitting< DATATYPE, IDXTYPE, INTERVALCONTAINER >.

◆ getNextSplit()

template<typename DATATYPE, typename IDXTYPE, class INTERVALCONTAINER>

bool FST::Data_Accessor_Splitting_Mem< DATATYPE, IDXTYPE, INTERVALCONTAINER >::getNextSplit ( )

overridevirtual

data access iteration (to support, e.g., loops in cross-validation)

Reimplemented from FST::Data_Accessor_Splitting< DATATYPE, IDXTYPE, INTERVALCONTAINER >.

Reimplemented in FST::Data_Accessor_Universe< DATATYPE, IDXTYPE, INTERVALCONTAINER >.

◆ initial_data_read()

template<typename DATATYPE, typename IDXTYPE, class INTERVALCONTAINER>

virtual void FST::Data_Accessor_Splitting_Mem< DATATYPE, IDXTYPE, INTERVALCONTAINER >::initial_data_read ( )

protectedpure virtual

fills the already-allocated data array with the file's samples; called once by initialize() after the array is sized Must write exactly datasize values, laid out class by class, pattern by pattern, feature by feature (class c starts at _class_offset[c]), matching the counts the file_get*() queries report.

Note: off-limits in shared_clone

Implemented in FST::Data_Accessor_Splitting_MemARFF< DATATYPE, IDXTYPE, INTERVALCONTAINER >, FST::Data_Accessor_Splitting_MemCSV< DATATYPE, IDXTYPE, INTERVALCONTAINER >, FST::Data_Accessor_Splitting_MemTRN< DATATYPE, IDXTYPE, INTERVALCONTAINER >, and FST::Data_Accessor_Universe< DATATYPE, IDXTYPE, INTERVALCONTAINER >.

◆ initial_data_scaling()

template<typename DATATYPE, typename IDXTYPE, class INTERVALCONTAINER>

void FST::Data_Accessor_Splitting_Mem< DATATYPE, IDXTYPE, INTERVALCONTAINER >::initial_data_scaling ( )

protected

Note: off-limits in shared_clone

◆ initial_file_prepare()

template<typename DATATYPE, typename IDXTYPE, class INTERVALCONTAINER>

virtual void FST::Data_Accessor_Splitting_Mem< DATATYPE, IDXTYPE, INTERVALCONTAINER >::initial_file_prepare ( )

protectedpure virtual

opens/prepares the underlying file for reading; called once by initialize() before initial_data_read() After it returns, the three file_get*() queries below must answer correctly.

Implemented in FST::Data_Accessor_Splitting_MemARFF< DATATYPE, IDXTYPE, INTERVALCONTAINER >, FST::Data_Accessor_Splitting_MemCSV< DATATYPE, IDXTYPE, INTERVALCONTAINER >, FST::Data_Accessor_Splitting_MemTRN< DATATYPE, IDXTYPE, INTERVALCONTAINER >, and FST::Data_Accessor_Universe< DATATYPE, IDXTYPE, INTERVALCONTAINER >.

◆ initialize()

template<typename DATATYPE, typename IDXTYPE, class INTERVALCONTAINER>

void FST::Data_Accessor_Splitting_Mem< DATATYPE, IDXTYPE, INTERVALCONTAINER >::initialize ( )

virtual

to be called after object creation (for initialization of data file access etc.)

Opens and reads the file through the derived class's hooks, allocates the in-memory data array and sets up the splitting structures. Call it exactly once, on the owning instance, before any other access; it throws fst_error (or an I/O or allocation exception) on failure. It also reports the file name, dimensionality, class sizes and splitting depth on standard output through syncout::print() – there is no quiet mode, so a caller that needs clean standard output must redirect it around this call.

Note: off-limits in shared_clone

Reimplemented in FST::Data_Accessor_Universe< DATATYPE, IDXTYPE, INTERVALCONTAINER >.

◆ print()

template<typename DATATYPE, typename IDXTYPE, class INTERVALCONTAINER>

std::ostream & FST::Data_Accessor_Splitting_Mem< DATATYPE, IDXTYPE, INTERVALCONTAINER >::print ( std::ostream & os ) const

overridevirtual

writes the splitter stack's print() output (or "Data_Accessor_Splitting()" if the stack is empty) to os

Reimplemented from FST::Data_Accessor_Splitting< DATATYPE, IDXTYPE, INTERVALCONTAINER >.

Reimplemented in FST::Data_Accessor_Splitting_MemARFF< DATATYPE, IDXTYPE, INTERVALCONTAINER >, FST::Data_Accessor_Splitting_MemCSV< DATATYPE, IDXTYPE, INTERVALCONTAINER >, FST::Data_Accessor_Splitting_MemTRN< DATATYPE, IDXTYPE, INTERVALCONTAINER >, and FST::Data_Accessor_Universe< DATATYPE, IDXTYPE, INTERVALCONTAINER >.

◆ scale_sample_inplace()

template<typename DATATYPE, typename IDXTYPE, class INTERVALCONTAINER>

void FST::Data_Accessor_Splitting_Mem< DATATYPE, IDXTYPE, INTERVALCONTAINER >::scale_sample_inplace ( DATATYPE * sample ) const

Applies the learned per-feature transform to sample, a caller-owned vector of getNoOfFeatures() values, in place.

Value i receives feature i's transform – min/max (to01), mean/stddev (white) or the missing-value substitution – exactly as the loaded data went through it, so a vector that is to be classified by a model trained on this accessor's data must pass through here first. In per-split mode (set_scaler_fit_train()) the transform is the one fitted on the CURRENT depth-0 split, i.e. the one the working buffer carries at the moment of the call. Throws fst_error before initialize(), or when the scaler does not implement duplicate().

◆ set_scaler_fit_train()

template<typename DATATYPE, typename IDXTYPE, class INTERVALCONTAINER>

void FST::Data_Accessor_Splitting_Mem< DATATYPE, IDXTYPE, INTERVALCONTAINER >::set_scaler_fit_train ( const bool b )

inline

Switches on per-split scaling: the scaler learns its transform on the current depth-0 split's TRAIN part only and applies it to ALL data, refreshed on every depth-0 split change.

This avoids the normalization leakage inherent in scaling at load time, where the test part contributes to the min/max or mean/stddev the transform is built from and so biases every later estimate of generalization performance. Must be called BEFORE initialize(). The default, false, scales once over the whole data set at load time.

Note: each instance (incl. every sharing_clone) then keeps a private working buffer and a private scaler instance, so threaded searches stay race-free and produce the same result as a sequential run

◆ sharing_clone()

template<typename DATATYPE, typename IDXTYPE, class INTERVALCONTAINER>

Data_Accessor_Splitting_Mem * FST::Data_Accessor_Splitting_Mem< DATATYPE, IDXTYPE, INTERVALCONTAINER >::sharing_clone ( ) const

overridepure virtual

create equivalent clone of the current object, permitting read-only access to structures in the source object (allows referencing instead of copying of large memory structures). may be faster and save space but requires more caution with respect to concurrency Use example: Data_Accessor memory data representation cloning

Implements FST::Data_Accessor< DATATYPE, IDXTYPE >.

◆ stateless_clone()

template<typename DATATYPE, typename IDXTYPE, class INTERVALCONTAINER>

Data_Accessor_Splitting_Mem * FST::Data_Accessor_Splitting_Mem< DATATYPE, IDXTYPE, INTERVALCONTAINER >::stateless_clone ( ) const

inlineoverridevirtual

create clone of the current object, ignoring internal temporary structures to save speed. Does not replicate exact object state. The clone must be used carefully in a way that ensures internal structure re-initialization Use example: Data_Splitter cloning or Classifier_LIBSVM cloning due to inability to clone external structures defined in LibSVM

Implements FST::Data_Accessor< DATATYPE, IDXTYPE >.

Member Data Documentation

◆ _class_offset

template<typename DATATYPE, typename IDXTYPE, class INTERVALCONTAINER>

std::vector<IDXTYPE> FST::Data_Accessor_Splitting_Mem< DATATYPE, IDXTYPE, INTERVALCONTAINER >::_class_offset

protected

class start indexes (offsets) in whatever random-access data representation

Note: read-only in shared_clone

◆ data

template<typename DATATYPE, typename IDXTYPE, class INTERVALCONTAINER>

std::shared_ptr<DATATYPE> FST::Data_Accessor_Splitting_Mem< DATATYPE, IDXTYPE, INTERVALCONTAINER >::data

protected

Note: read-only in shared_clone — EXCEPT in scaler_fit_train mode, where each instance owns a private working buffer

◆ dsc

template<typename DATATYPE, typename IDXTYPE, class INTERVALCONTAINER>

PScaler FST::Data_Accessor_Splitting_Mem< DATATYPE, IDXTYPE, INTERVALCONTAINER >::dsc

protected

Note: in scaler_fit_train mode every instance holds a private duplicate(); otherwise off-limits in shared_clone

◆ feature_scalers

template<typename DATATYPE, typename IDXTYPE, class INTERVALCONTAINER>

std::vector<PScaler> FST::Data_Accessor_Splitting_Mem< DATATYPE, IDXTYPE, INTERVALCONTAINER >::feature_scalers

protected

the learned per-feature transforms: one duplicate() of dsc per feature, taken right after that feature was learned, for scale_sample_inplace() – the single dsc object is re-learned feature after feature and so ends up holding only the LAST feature's statistics. Filled by initial_data_scaling() (load-time fit), or by per_split_data_scaling() in scaler_fit_train mode (then the current depth-0 split's transform); empty until then, or when the scaler does not implement duplicate(). The duplicates are never mutated, so sharing clones may share them.

◆ filename

template<typename DATATYPE, typename IDXTYPE, class INTERVALCONTAINER>

std::string FST::Data_Accessor_Splitting_Mem< DATATYPE, IDXTYPE, INTERVALCONTAINER >::filename

protected

Note: off-limits in shared_clone

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

/Users/psms/dev/devtest1/_src_dataio/data_accessor_splitting_mem.hpp

Public Types

Public Member Functions

Protected Member Functions

Protected Attributes

Additional Inherited Members

Detailed Description

Member Typedef Documentation

◆ PSplitters

Constructor & Destructor Documentation

◆ Data_Accessor_Splitting_Mem()

Member Function Documentation

◆ clone()

◆ file_getClassSize()

◆ file_getNoOfClasses()

◆ file_getNoOfFeatures()

◆ getFirstBlock()

◆ getFirstSplit()

◆ getNextBlock()

◆ getNextSplit()

◆ initial_data_read()

◆ initial_data_scaling()

◆ initial_file_prepare()

◆ initialize()

◆ print()

◆ scale_sample_inplace()

◆ set_scaler_fit_train()

◆ sharing_clone()

◆ stateless_clone()

Member Data Documentation

◆ _class_offset

◆ data

◆ dsc

◆ feature_scalers

◆ filename