#include <toolkits/factorization/ranking_sgd_solver_base.hpp>

Classes
struct	neg_sample_proc_buffer

Public Member Functions
std::pair< double, double >	run_iteration (size_t iteration, sgd::sgd_interface_base *model_iface, const v2::ml_data &data, double step_size)

std::pair< double, double >	calculate_objective (sgd::sgd_interface_base *model_iface, const v2::ml_data &data, size_t iteration) const GL_HOT

std::map< std::string, variant_type >	run ()

Protected Member Functions
	ranking_sgd_solver_base (const std::shared_ptr< sgd::sgd_interface_base > &main_interface, const v2::ml_data &train_data, const std::map< std::string, flexible_type > &options)

virtual std::pair< double, double >	run_sgd_thread (size_t iteration, size_t thread_idx, size_t num_threads, size_t block_idx, size_t num_blocks, const v2::ml_data &data, SGDInterface *iface, double step_size, volatile bool &error_detected)=0

virtual std::pair< double, double >	run_loss_calculation_thread (size_t thread_idx, size_t num_threads, const v2::ml_data &data, SGDInterface *iface) const =0

std::pair< size_t, size_t >	fill_x_buffer_with_users_items (std::vector< std::pair< std::vector< v2::ml_data_entry >, double > > &x_buffer, v2::ml_data_block_iterator &it, size_t n_items, dense_bitset &item_observed) const GL_HOT_INLINE_FLATTEN

template<typename BufferIndexToItemIndexMapper >
GL_HOT_INLINE_FLATTEN void	clear_item_observed_buffer (dense_bitset &item_observed, size_t n_rows, size_t n_items, const BufferIndexToItemIndexMapper &map_index) const

virtual void	setup (sgd_interface_base *iface)

Static Protected Member Functions
static void	add_options (option_manager &options)

Protected Attributes
std::shared_ptr< sgd_interface_base >	model_interface

const std::map< std::string, flexible_type >	options

Detailed Description

template<class SGDInterface>
class turi::factorization::ranking_sgd_solver_base< SGDInterface >

The main parts of the ranking sgd solver class. Part of the functionality is implemented in one of two subclasses, explicit_ranking_sgd_solver and implicit_ranking_sgd_solver.

The ranking SGD stuff is broken into two categories; ranking regularization alongside training to predict a target, and "pure" ranking in which there is no target. Both require a substantial amount of bookkeeping to handle the negative item examples. The details of how these are applied, however, is handled with the individual subclasses.

Definition at line 33 of file ranking_sgd_solver_base.hpp.

Constructor & Destructor Documentation

◆ ranking_sgd_solver_base()

template<class SGDInterface >

turi::factorization::ranking_sgd_solver_base< SGDInterface >::ranking_sgd_solver_base	(	const std::shared_ptr< sgd::sgd_interface_base > &	main_interface,
		const v2::ml_data &	train_data,
		const std::map< std::string, flexible_type > &	options
	)

inlineprotected

Constructor

Definition at line 45 of file ranking_sgd_solver_base.hpp.

Member Function Documentation

◆ add_options()

static void turi::sgd::sgd_solver_base::add_options ( option_manager & options )

staticprotectedinherited

Call the following function to insert the option definitions needed for the common sgd optimization class into an option manager. Meant to be called by the subclasses of sgd_solver_base.

◆ calculate_objective()

template<class SGDInterface >

std::pair<double, double> turi::factorization::ranking_sgd_solver_base< SGDInterface >::calculate_objective	(	sgd::sgd_interface_base *	model_iface,
		const v2::ml_data &	data,
		size_t	iteration
	)		const

inlinevirtual

Calculate the objective value of the current state.

Parameters

[in]	iface_base	The interface class that gives the gradient calculation routines on top of the model definition. This must be upcast to SGDInterface*.
[in]	row_start	The starting row in the training data to use. In trial mode, we are likely looking at only a subset of the data.
[in]	row_end	The ending row in the training data to use. In trial mode, we are likely looking at only a subset of the data.

Reimplemented from turi::sgd::sgd_solver_base.

Definition at line 277 of file ranking_sgd_solver_base.hpp.

◆ clear_item_observed_buffer()

template<class SGDInterface >

template<typename BufferIndexToItemIndexMapper >

GL_HOT_INLINE_FLATTEN void turi::factorization::ranking_sgd_solver_base< SGDInterface >::clear_item_observed_buffer	(	dense_bitset &	item_observed,
		size_t	n_rows,
		size_t	n_items,
		const BufferIndexToItemIndexMapper &	map_index
	)		const

inlineprotected

Clear out the item_observed buffer.

Based on the number of items actually used, deletes stuff. Defined below.

Definition at line 656 of file ranking_sgd_solver_base.hpp.

◆ fill_x_buffer_with_users_items()

template<class SGDInterface >

std::pair<size_t, size_t> turi::factorization::ranking_sgd_solver_base< SGDInterface >::fill_x_buffer_with_users_items	(	std::vector< std::pair< std::vector< v2::ml_data_entry >, double > > &	x_buffer,
		v2::ml_data_block_iterator &	it,
		size_t	n_items,
		dense_bitset &	item_observed
	)		const

inlineprotected

Fill a buffer with (observation, target value) pairs. Because of the user-block nature of the ml_data_block_iterator, this buffer is gauranteed to hold all the items rated by a particular user. If no target_value is present, then "1" is used.

Parameters

[out]	x_buffer	The buffer where the (observation, target_value) pairs are stored.
[in,out]	it	The current block iterator.
[in,out]	item_observed	A mask giving the items observed in the data.

Returns: (n_rows, n_rated_items). The number of rows in the buffer, and the number of unique rated items.

Definition at line 360 of file ranking_sgd_solver_base.hpp.

◆ run()

std::map<std::string, variant_type> turi::sgd::sgd_solver_base::run ( )

inherited

The main function to run the sgd solver given the current options.

◆ run_iteration()

template<class SGDInterface >

std::pair<double, double> turi::factorization::ranking_sgd_solver_base< SGDInterface >::run_iteration	(	size_t	iteration,
		sgd::sgd_interface_base *	model_iface,
		const v2::ml_data &	data,
		double	step_size
	)

inlinevirtual

Run a single SGD pass through the data. Implementation of base sgd_solver's required virtual function.

Parameters

[in]	iteration	The iteration index; what gets reported in the progress message.
[in]	iface_base	The interface class that gives the gradient calculation routines on top of the model definition. This must be upcast to SGDInterface*.
[in]	row_start	The starting row in the training data to use. In trial mode, we are likely looking at only a subset of the data.
[in]	row_end	The ending row in the training data to use. In trial mode, we are likely looking at only a subset of the data.
[in]	trial_mode	If true, immediately return failure on any numerical issues and do not report progress messages.

Implements turi::sgd::sgd_solver_base.

Definition at line 158 of file ranking_sgd_solver_base.hpp.

◆ run_loss_calculation_thread()

template<class SGDInterface >

virtual std::pair<double, double> turi::factorization::ranking_sgd_solver_base< SGDInterface >::run_loss_calculation_thread	(	size_t	thread_idx,
		size_t	num_threads,
		const v2::ml_data &	data,
		SGDInterface *	iface
	)		const

protectedpure virtual

Calculate the loss value for the block of data assigned to a particular thread.

Parameters

[in]	thread_idx	The thread index determining this block.
[in]	num_threads	The number of threads.
[in]	data	The ml_data instance we're working with. Primarily needed for the metadata.
[in]	it_init	The iterator inializer for the ml_data_block_iterator used for this thread.
[in]	iface	The working SGD interface.

Returns: (loss, rank_loss) – loss is the cumulative estimated loss value for this thread on predicting the training data, and rank_loss is the weighted loss on the negative examples.

◆ run_sgd_thread()

template<class SGDInterface >

virtual std::pair<double, double> turi::factorization::ranking_sgd_solver_base< SGDInterface >::run_sgd_thread	(	size_t	iteration,
		size_t	thread_idx,
		size_t	num_threads,
		size_t	block_idx,
		size_t	num_blocks,
		const v2::ml_data &	data,
		SGDInterface *	iface,
		double	step_size,
		volatile bool &	error_detected
	)

protectedpure virtual

The main method needed to be implemented by the subclass to run the ranking sgd stuff.

Parameters

[in]	thread_idx	The thread index determining this block.
[in]	num_threads	The number of threads.
[in]	data	The v2::ml_data instance we're working with. Primarily needed for the metadata.
[in]	it_init	The iterator inializer for the ml_data_block_iterator used for this thread.
[in]	iface	The working SGD interface.
[in]	step_size	The current SGD step size, set by the higher level algorithm.
[in,out]	error_detected	If set to true, a numerical error is detected.

Returns: (loss, rank_loss) – loss is the cumulative estimated loss value for this thread on predicting the training data, and rank_loss is the weighted loss on the negative examples.

◆ setup()

virtual void turi::sgd::sgd_solver_base::setup ( sgd_interface_base * iface )

inlineprotectedvirtualinherited

Called at the start of a run, before any run_iteration is called.

Definition at line 109 of file sgd_solver_base.hpp.

Member Data Documentation

◆ model_interface

std::shared_ptr<sgd_interface_base> turi::sgd::sgd_solver_base::model_interface

protectedinherited

The main interface to the model, implementing sgd-specific routines for that model.

Definition at line 96 of file sgd_solver_base.hpp.

◆ options

const std::map<std::string, flexible_type> turi::sgd::sgd_solver_base::options

protectedinherited

The training options of the solver.

Definition at line 100 of file sgd_solver_base.hpp.

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

toolkits/factorization/ranking_sgd_solver_base.hpp

Classes

Public Member Functions

Protected Member Functions

Static Protected Member Functions

Protected Attributes

Detailed Description

template<class SGDInterface> class turi::factorization::ranking_sgd_solver_base< SGDInterface >

Constructor & Destructor Documentation

◆ ranking_sgd_solver_base()

Member Function Documentation

◆ add_options()

◆ calculate_objective()

◆ clear_item_observed_buffer()

◆ fill_x_buffer_with_users_items()

◆ run()

◆ run_iteration()

◆ run_loss_calculation_thread()

◆ run_sgd_thread()

◆ setup()

Member Data Documentation

◆ model_interface

◆ options

template<class SGDInterface>
class turi::factorization::ranking_sgd_solver_base< SGDInterface >