Sparse Linear Methods (SLIM) for item prediction (ranking) optimized for BPR-Opt optimization criterion More...

Inheritance diagram for BPRSLIM:

Public Member Functions
virtual void	AddFeedback (ICollection< Tuple< int, int >> feedback)
	Add positive feedback events and perform incremental training More...

	BPRSLIM ()
	Default constructor More...

virtual bool	CanPredict (int user_id, int item_id)
	Check whether a useful prediction (i.e. not using a fallback/default answer) can be made for a given user-item combination More...

Object	Clone ()
	create a shallow copy of the object More...

override float	ComputeObjective ()
	Compute the fit (AUC on training data) More...

override void	Iterate ()
	Perform one iteration of stochastic gradient ascent over the training data More...

override void	LoadModel (string file)
	Get the model parameters from a file More...

override float	Predict (int user_id, int item_id)
	Predict rating or score for a given user-item combination More...

double	PredictWithDifference (int user_id, int pos_item_id, int neg_item_id)

IList< Tuple< int, float > >	Recommend (int user_id, int n=-1, ICollection< int > ignore_items=null, ICollection< int > candidate_items=null)
	Recommend items for a given user More...

virtual System.Collections.Generic.IList< Tuple< int, float > >	Recommend (int user_id, int n=-1, System.Collections.Generic.ICollection< int > ignore_items=null, System.Collections.Generic.ICollection< int > candidate_items=null)

virtual void	RemoveFeedback (ICollection< Tuple< int, int >> feedback)
	Remove all feedback events by the given user-item combinations More...

override void	RemoveItem (int item_id)
	Remove all feedback by one item More...

virtual void	RemoveUser (int user_id)
	Remove all feedback by one user More...

override void	SaveModel (string file)
	Save the model parameters to a file More...

override string	ToString ()
	Return a string representation of the recommender More...

override void	Train ()
	Learn the model parameters of the recommender from the training data More...

Protected Member Functions
override void	AddItem (int item_id)

override void	AddUser (int user_id)

override void	InitModel ()

virtual void	RetrainItem (int item_id)
	Retrain the latent factors of a given item More...

virtual void	SampleItemPair (int u, out int i, out int j)
	Sample a pair of items, given a user More...

virtual bool	SampleOtherItem (int u, int i, out int j)
	Sample another item, given the first one and the user More...

virtual void	SampleTriple (out int u, out int i, out int j)
	Sample a triple for BPR learning More...

virtual int	SampleUser ()
	Sample a user that has viewed at least one and not all items More...

virtual void	UpdateFactors (int u, int i, int j, bool update_u, bool update_i, bool update_j)
	Update latent factors according to the stochastic gradient descent update rule More...

Protected Attributes
Matrix< float >	item_weights
	Item weight matrix (the W matrix in the original paper) More...

ItemKNN	itemKNN
	The item KNN used in the feature selection step More...

double	learn_rate = 0.05
	Learning rate alpha More...

System.Random	random
	Random number generator More...

double	reg_i = 0.0025
	Regularization parameter for positive item weights More...

double	reg_j = 0.00025
	Regularization parameter for negative item weights More...

bool	update_j = true
	If set (default), update factors for negative sampled items during learning More...

Properties
virtual IPosOnlyFeedback	Feedback `[get, set]`
	the feedback data to be used for training More...

double	InitMean `[get, set]`
	Mean of the normal distribution used to initialize the latent factors More...

double	InitStdDev `[get, set]`
	Standard deviation of the normal distribution used to initialize the latent factors More...

double	LearnRate `[get, set]`
	Learning rate alpha More...

int	MaxItemID `[get, set]`
	Maximum item ID More...

int	MaxUserID `[get, set]`
	Maximum user ID More...

uint	NumIter `[get, set]`
	Number of iterations over the training data More...

double	RegI `[get, set]`
	Regularization parameter for positive item weights More...

double	RegJ `[get, set]`
	Regularization parameter for negative item weights More...

bool	UniformUserSampling `[get, set]`
	Sample uniformly from users More...

bool	UpdateItems `[get, set]`

bool	UpdateJ `[get, set]`
	If set (default), update factors for negative sampled items during learning More...

bool	UpdateUsers `[get, set]`

bool	WithReplacement `[get, set]`
	Sample positive observations with (true) or without (false) replacement More...

Detailed Description

Sparse Linear Methods (SLIM) for item prediction (ranking) optimized for BPR-Opt optimization criterion

This implementation differs from the algorithm in the original SLIM paper since the model here is optimized for BPR-Opt instead of the elastic net loss. The optmization algorithm used is the Sotchastic Gradient Ascent.

Literature:

Steffen Rendle, Christoph Freudenthaler, Zeno Gantner, Lars Schmidt-Thieme: BPR: Bayesian Personalized Ranking from Implicit Feedback. UAI 2009. http://www.ismll.uni-hildesheim.de/pub/pdfs/Rendle_et_al2009-Bayesian_Personalized_Ranking.pdf
X. Ning, G. Karypis: Slim: Sparse linear methods for top-n recommender systems. ICDM 2011. http://glaros.dtc.umn.edu/gkhome/fetch/papers/SLIM2011icdm.pdf

Different sampling strategies are configurable by setting the UniformUserSampling and WithReplacement accordingly. To get the strategy from the original paper, set UniformUserSampling=false and WithReplacement=false. WithReplacement=true (default) gives you usually a slightly faster convergence, and UniformUserSampling=true (default) (approximately) optimizes the average AUC over all users.

This recommender supports incremental updates.

Constructor & Destructor Documentation

BPRSLIM ( )

inline

Default constructor

Member Function Documentation

virtual void AddFeedback ( ICollection< Tuple< int, int >> feedback )

inlinevirtualinherited

Add positive feedback events and perform incremental training

Parameters

feedback collection of user id - item id tuples

Implements IIncrementalItemRecommender.

Reimplemented in UserKNN, ItemKNN, MostPopular, and MF.

virtual bool CanPredict	(	int	user_id,
		int	item_id
	)

inlinevirtualinherited

Check whether a useful prediction (i.e. not using a fallback/default answer) can be made for a given user-item combination

It is up to the recommender implementor to decide when a prediction is useful, and to document it accordingly.

Parameters

user_id	the user ID
item_id	the item ID

Returns: true if a useful prediction can be made, false otherwise

Implements IRecommender.

Reimplemented in ExternalItemRecommender, ExternalRatingPredictor, BiPolarSlopeOne, SlopeOne, Constant, GlobalAverage, UserAverage, ItemAverage, and Random.

Object Clone ( )

inlineinherited

create a shallow copy of the object

override float ComputeObjective ( )

inlinevirtual

Compute the fit (AUC on training data)

Returns: the fit

Implements SLIM.

override void Iterate ( )

inlinevirtual

Perform one iteration of stochastic gradient ascent over the training data

One iteration is samples number of positive entries in the training matrix times

Implements SLIM.

override void LoadModel ( string filename )

inlinevirtual

Get the model parameters from a file

Parameters

filename the name of the file to read from

Reimplemented from Recommender.

override float Predict	(	int	user_id,
		int	item_id
	)

inlinevirtualinherited

Predict rating or score for a given user-item combination

Parameters

user_id	the user ID
item_id	the item ID

Returns: the predicted score/rating for the given user-item combination

Implements Recommender.

IList<Tuple<int, float> > Recommend	(	int	user_id,
		int	n = `-1`,
		ICollection< int >	ignore_items = `null`,
		ICollection< int >	candidate_items = `null`
	)

inherited

Recommend items for a given user

Parameters

user_id	the user ID
n	the number of items to recommend, -1 for as many as possible
ignore_items	collection if items that should not be returned; if null, use empty collection
candidate_items	the candidate items to choose from; if null, use all items

Returns: a sorted list of (item_id, score) tuples

Implemented in WeightedEnsemble, and Ensemble.

virtual void RemoveFeedback ( ICollection< Tuple< int, int >> feedback )

inlinevirtualinherited

Remove all feedback events by the given user-item combinations

Parameters

feedback collection of user id - item id tuples

Implements IIncrementalItemRecommender.

Reimplemented in UserKNN, MostPopular, ItemKNN, and MF.

override void RemoveItem ( int item_id )

inlinevirtual

Remove all feedback by one item

Parameters

item_id the item ID

Reimplemented from IncrementalItemRecommender.

virtual void RemoveUser ( int user_id )

inlinevirtualinherited

Remove all feedback by one user

Parameters

user_id the user ID

Implements IIncrementalRecommender.

Reimplemented in LeastSquareSLIM, MF, and MostPopular.

virtual void RetrainItem ( int item_id )

inlineprotectedvirtual

Retrain the latent factors of a given item

Parameters

item_id the item ID

virtual void SampleItemPair	(	int	u,
		out int	i,
		out int	j
	)

inlineprotectedvirtual

Sample a pair of items, given a user

Parameters

u	the user ID
i	the ID of the first item
j	the ID of the second item

virtual bool SampleOtherItem	(	int	u,
		int	i,
		out int	j
	)

inlineprotectedvirtual

Sample another item, given the first one and the user

Parameters

u	the user ID
i	the ID of the given item
j	the ID of the other item

Returns: true if the given item was already seen by user u

virtual void SampleTriple	(	out int	u,
		out int	i,
		out int	j
	)

inlineprotectedvirtual

Sample a triple for BPR learning

Parameters

u	the user ID
i	the ID of the first item
j	the ID of the second item

virtual int SampleUser ( )

inlineprotectedvirtual

Sample a user that has viewed at least one and not all items

Returns: the user ID

override void SaveModel ( string filename )

inlinevirtual

Save the model parameters to a file

Parameters

filename the name of the file to write to

Reimplemented from Recommender.

override string ToString ( )

inline

Return a string representation of the recommender

The ToString() method of recommenders should list the class name and all hyperparameters, separated by space characters.

Implements IRecommender.

override void Train ( )

inlinevirtual

Learn the model parameters of the recommender from the training data

Implements Recommender.

virtual void UpdateFactors	(	int	u,
		int	i,
		int	j,
		bool	update_u,
		bool	update_i,
		bool	update_j
	)

inlineprotectedvirtual

Update latent factors according to the stochastic gradient descent update rule

Parameters

u	the user ID
i	the ID of the first item
j	the ID of the second item
update_u	if true, update the user latent factors
update_i	if true, update the latent factors of the first item
update_j	if true, update the latent factors of the second item

Member Data Documentation

Matrix<float> item_weights

protectedinherited

Item weight matrix (the W matrix in the original paper)

ItemKNN itemKNN

protectedinherited

The item KNN used in the feature selection step

double learn_rate = 0.05

protected

Learning rate alpha

System.Random random

protected

Random number generator

double reg_i = 0.0025

protected

Regularization parameter for positive item weights

double reg_j = 0.00025

protected

Regularization parameter for negative item weights

bool update_j = true

protected

If set (default), update factors for negative sampled items during learning

Property Documentation

virtual IPosOnlyFeedback Feedback

getsetinherited

the feedback data to be used for training

double InitMean

getsetinherited

Mean of the normal distribution used to initialize the latent factors

double InitStdDev

getsetinherited

Standard deviation of the normal distribution used to initialize the latent factors

double LearnRate

getset

Learning rate alpha

int MaxItemID

getsetinherited

Maximum item ID

int MaxUserID

getsetinherited

Maximum user ID

uint NumIter

getsetinherited

Number of iterations over the training data

double RegI

getset

Regularization parameter for positive item weights

double RegJ

getset

Regularization parameter for negative item weights

bool UniformUserSampling

getset

Sample uniformly from users

bool UpdateJ

getset

If set (default), update factors for negative sampled items during learning

bool WithReplacement

getset

Sample positive observations with (true) or without (false) replacement

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

BPRSLIM.cs

Public Member Functions

Protected Member Functions

Protected Attributes

Properties

Detailed Description

Constructor & Destructor Documentation

Member Function Documentation

Member Data Documentation

Property Documentation