turi::statistics_tracker Class Reference

#include <toolkits/feature_engineering/statistics_tracker.hpp>

Public Member Functions
	statistics_tracker (const std::string _column_name="")
	statistics_tracker (const statistics_tracker &)=delete
void	initialize ()
void	insert_or_update (const flexible_type &key, flexible_type value, size_t thread_idx=0) GL_HOT
size_t	lookup (const flexible_type &value) const
size_t	lookup_counts (const flexible_type &value) const
flex_float	lookup_means (const flexible_type &value) const
void	finalize (size_t num_examples)
flexible_type	inverse_lookup (size_t idx) const
size_t	size () const
std::vector< flexible_type >	get_keys () const
size_t	get_version () const
void	save_impl (turi::oarchive &oarc) const
void	load_version (turi::iarchive &iarc, size_t version)

Detailed Description

Parallel statistics(mean) tracker

Note: This implementation is intended to be general and will be moved to some place more general later.

Construction

// Construct the tracker with the arguments. auto tracker = statistics_tracker(10, 1, "column_name_for_error_messages"); tracker.initialize();

// Insert flexible types into the tracker for (const flexible_type& v: sa.range_iterator() { tracker.insert_or_update(v); }

// Finalize mapping tracker.finalize();

Lookups

size_t index = tracker.lookup(v); // Returns (size_t) -1 if not present.

size_t counts = tracker.lookup_counts(v); // Returns 0 if not present.

flexible_type v = tracker.inverse_lookup(1) // Fails if index doesn't exist.

Parallel construction

// Initialize tracker.initialize();

// Perform the indexing. in_parallel([&](size_t thread_idx, size_t num_threads) {

size_t start_idx = src_size * thread_idx / num_threads; size_t end_idx = src_size * (thread_idx + 1) / num_threads;

// Inserts value of 1 for each key k for (const flexible_type& k: sa.range_iterator(start_idx, end_idx) { tracker.insert_or_update(k,1,thread_id); }

// Finalize tracker.finalize();

Definition at line 72 of file statistics_tracker.hpp.

Constructor & Destructor Documentation

statistics_tracker() [1/2]

turi::statistics_tracker::statistics_tracker ( const std::string  _column_name = "" )

inline

Default constructor

Parameters

[in]

column_name

Column name for display.

Definition at line 82 of file statistics_tracker.hpp.

statistics_tracker() [2/2]

turi::statistics_tracker::statistics_tracker ( const statistics_tracker &  )

delete

Copy constructor: Don't want to risk making copies of this.

Member Function Documentation

finalize()

void turi::statistics_tracker::finalize

(

size_t

num_examples

)

Finalize by dropping indices that dont meet

• Count requirement i.e count >= threshold.
• Topk requirement.

get_keys()

std::vector<flexible_type> turi::statistics_tracker::get_keys ( ) const

inline

Returns the number of categorical variables.

Returns

Column size.

Definition at line 162 of file statistics_tracker.hpp.

get_version()

size_t turi::statistics_tracker::get_version

(

)

const

Returns the current version used for the serialization.

initialize()

void turi::statistics_tracker::initialize

(

)

Initialize the index mapping and setup. Should be called before starting the map.

insert_or_update()

void turi::statistics_tracker::insert_or_update	(	const flexible_type &	key,
		flexible_type	value,
		size_t	thread_idx = 0
	)

Insert

Parameters

[in]	key	Flexible type.
[in]	value	Flexible type.
[in]	thread_idx	Thread id (For parallel insertion).

inverse_lookup()

flexible_type turi::statistics_tracker::inverse_lookup

(

size_t

idx

)

const

Returns the "value" associated with the index.

Parameters

load_version()

void turi::statistics_tracker::load_version	(	turi::iarchive &	iarc,
		size_t	version
	)

Load the object.

lookup()

size_t turi::statistics_tracker::lookup

(

const flexible_type &

value

)

const

Returns the index associated with the value.

Parameters

[in]

value

Search for the value.

Returns

The index. (Returns size_t(-1) if not present).

lookup_counts()

size_t turi::statistics_tracker::lookup_counts

(

const flexible_type &

value

)

const

Returns the counts associated with the value.

Parameters

[in]

value

Search for the value.

Returns

Counts (Returns 0 if not present).

lookup_means()

flex_float turi::statistics_tracker::lookup_means

(

const flexible_type &

value

)

const

Returns the counts associated with the value.

Parameters

[in]

value

Search for the value.

Returns

Counts (Returns 0 if not present).

save_impl()

void turi::statistics_tracker::save_impl

(

turi::oarchive &

oarc

)

const

Serialize the object (save).

size()

size_t turi::statistics_tracker::size ( ) const

inline

Returns the number of categorical variables.

Returns

Column size.

Definition at line 154 of file statistics_tracker.hpp.

---

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

• toolkits/feature_engineering/statistics_tracker.hpp