io.kgraph.library.cf

Class Svdpp

java.lang.Object

io.kgraph.library.cf.Svdpp

All Implemented Interfaces:

ComputeFunction<CfLongId,Svdpp.SvdppValue,Float,FloatMatrixMessage>

public class Svdpp
extends Object
implements ComputeFunction<CfLongId,Svdpp.SvdppValue,Float,FloatMatrixMessage>

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
class	Svdpp.InitItemsComputation
class	Svdpp.InitUsersComputation This computation class is used to initialize the factors of the user nodes in the very first superstep, and send the first updates to the item nodes.
class	Svdpp.ItemComputation
static class	Svdpp.SvdppValue A value in the Svdpp algorithm consists of (i) the baseline estimate, (ii) the latent vector, and (iii) the weight vector.
class	Svdpp.UserComputation

Nested classes/interfaces inherited from interface io.kgraph.pregel.ComputeFunction

ComputeFunction.Aggregators, ComputeFunction.Callback<K,VV,EV,Message>, ComputeFunction.InitCallback, ComputeFunction.MasterCallback, ComputeFunction.ReadAggregators, ComputeFunction.ReadWriteAggregators

Field Summary

Fields

Modifier and Type	Field and Description
static String	BIAS_GAMMA Bias learning rate parameter
static float	BIAS_GAMMA_DEFAULT Default value for bias learning rate
static String	BIAS_LAMBDA Bias regularization parameter.
static float	BIAS_LAMBDA_DEFAULT Default value for bias regularization parameter
static String	FACTOR_GAMMA Factor learning rate parameter
static float	FACTOR_GAMMA_DEFAULT Default value for factor learning rate
static String	FACTOR_LAMBDA Factor regularization parameter.
static float	FACTOR_LAMBDA_DEFAULT Default value for factor regularization parameter
static String	ITERATIONS Maximum number of iterations.
static int	ITERATIONS_DEFAULT Default value for ITERATIONS.
static String	MAX_RATING Max rating.
static float	MAX_RATING_DEFAULT Default maximum rating
static String	MIN_RATING Min rating.
static float	MIN_RATING_DEFAULT Default minimum rating
static String	OVERALL_RATING_AGGREGATOR Name of aggregator that aggregates all ratings.
static String	RANDOM_SEED Random seed.
static Long	RANDOM_SEED_DEFAULT Default random seed
static String	RMSE_AGGREGATOR Aggregator for the computation of RMSE
static String	RMSE_TARGET RMSE target
static float	RMSE_TARGET_DEFAULT Default value for parameter enabling the RMSE aggregator.
static String	VECTOR_SIZE Latent vector size.
static int	VECTOR_SIZE_DEFAULT Default latent vector size

Constructor Summary

Constructors

Constructor and Description
Svdpp()

Method Summary

Modifier and Type	Method and Description
void	compute(int superstep, VertexWithValue<CfLongId,Svdpp.SvdppValue> vertex, Iterable<FloatMatrixMessage> messages, Iterable<EdgeWithValue<CfLongId,Float>> edges, ComputeFunction.Callback<CfLongId,Svdpp.SvdppValue,Float,FloatMatrixMessage> cb) The function for computing a new vertex value or sending messages to the next superstep.
protected static float	computePredictedRating(float meanRating, float userBaseline, float itemBaseline, org.jblas.FloatMatrix user, org.jblas.FloatMatrix item, int numRatings, org.jblas.FloatMatrix sumWeights, float minRating, float maxRating) Computes the predicted rating r between a user and an item based on the formula: r = b + q^T * (p + (1/sqrt(N) * sum(y_i)))
protected static float	computeUpdatedBaseLine(float baseline, float predictedRating, float observedRating, float gamma, float lambda) Computes the updated baseline based on the formula:
protected long	getTotalNumEdges(ComputeFunction.ReadAggregators aggregators)
protected static float	incrementValue(float baseline, float step, float gamma, float lambda) Increments a scalar value according to the formula:
protected static void	incrementValue(org.jblas.FloatMatrix value, org.jblas.FloatMatrix step, float gamma, float lambda) Increments a vector according to the formula
void	init(Map<String,?> configs, ComputeFunction.InitCallback cb) Initialize the ComputeFunction, this is the place to register aggregators.
void	masterCompute(int superstep, ComputeFunction.MasterCallback cb) A function for performing sequential computations between supersteps.
void	preSuperstep(int superstep, ComputeFunction.Aggregators aggregators) Prepare for computation.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Methods inherited from interface io.kgraph.pregel.ComputeFunction

postSuperstep

Field Detail

OVERALL_RATING_AGGREGATOR

public static final String OVERALL_RATING_AGGREGATOR

Name of aggregator that aggregates all ratings.

See Also:

Constant Field Values

RMSE_TARGET

public static final String RMSE_TARGET

RMSE target

See Also:

Constant Field Values

RMSE_TARGET_DEFAULT

public static final float RMSE_TARGET_DEFAULT

Default value for parameter enabling the RMSE aggregator.

See Also:

Constant Field Values

ITERATIONS

public static final String ITERATIONS

Maximum number of iterations.

See Also:

Constant Field Values

ITERATIONS_DEFAULT

public static final int ITERATIONS_DEFAULT

Default value for ITERATIONS.

See Also:

Constant Field Values

FACTOR_LAMBDA

public static final String FACTOR_LAMBDA

Factor regularization parameter.

See Also:

Constant Field Values

FACTOR_LAMBDA_DEFAULT

public static final float FACTOR_LAMBDA_DEFAULT

Default value for factor regularization parameter

See Also:

Constant Field Values

FACTOR_GAMMA

public static final String FACTOR_GAMMA

Factor learning rate parameter

See Also:

Constant Field Values

FACTOR_GAMMA_DEFAULT

public static final float FACTOR_GAMMA_DEFAULT

Default value for factor learning rate

See Also:

Constant Field Values

BIAS_LAMBDA

public static final String BIAS_LAMBDA

Bias regularization parameter.

See Also:

Constant Field Values

BIAS_LAMBDA_DEFAULT

public static final float BIAS_LAMBDA_DEFAULT

Default value for bias regularization parameter

See Also:

Constant Field Values

BIAS_GAMMA

public static final String BIAS_GAMMA

Bias learning rate parameter

See Also:

Constant Field Values

BIAS_GAMMA_DEFAULT

public static final float BIAS_GAMMA_DEFAULT

Default value for bias learning rate

See Also:

Constant Field Values

MAX_RATING

public static final String MAX_RATING

Max rating.

See Also:

Constant Field Values

MAX_RATING_DEFAULT

public static final float MAX_RATING_DEFAULT

Default maximum rating

See Also:

Constant Field Values

MIN_RATING

public static final String MIN_RATING

Min rating.

See Also:

Constant Field Values

MIN_RATING_DEFAULT

public static final float MIN_RATING_DEFAULT

Default minimum rating

See Also:

Constant Field Values

VECTOR_SIZE

public static final String VECTOR_SIZE

Latent vector size.

See Also:

Constant Field Values

VECTOR_SIZE_DEFAULT

public static final int VECTOR_SIZE_DEFAULT

Default latent vector size

See Also:

Constant Field Values

RANDOM_SEED

public static final String RANDOM_SEED

Random seed.

See Also:

Constant Field Values

RANDOM_SEED_DEFAULT

public static final Long RANDOM_SEED_DEFAULT

Default random seed

RMSE_AGGREGATOR

public static final String RMSE_AGGREGATOR

Aggregator for the computation of RMSE

See Also:

Constant Field Values

Constructor Detail

Svdpp

public Svdpp()

Method Detail

computePredictedRating

protected static float computePredictedRating(float meanRating,
                                              float userBaseline,
                                              float itemBaseline,
                                              org.jblas.FloatMatrix user,
                                              org.jblas.FloatMatrix item,
                                              int numRatings,
                                              org.jblas.FloatMatrix sumWeights,
                                              float minRating,
                                              float maxRating)

Computes the predicted rating r between a user and an item based on the formula: r = b + q^T * (p + (1/sqrt(N) * sum(y_i)))

where b: the baseline estimate of the user for the item q: the item vector p: the user vector N: number of ratings of the user y_i: the weight vector

Parameters:

meanRating - the mean rating

userBaseline - the user baseline

itemBaseline - the item baseline

user - the user

item - the item

numRatings - the number of ratings

sumWeights - the sum of weights

maxRating - the max rating

minRating - the min rating

Returns:

the predicted rating

computeUpdatedBaseLine

protected static float computeUpdatedBaseLine(float baseline,
                                              float predictedRating,
                                              float observedRating,
                                              float gamma,
                                              float lambda)

Computes the updated baseline based on the formula:

b := b + gamma * (error - lambda * b)

Parameters:

baseline - the baseline

predictedRating - the predicted rating

observedRating - the observed rating

gamma - the gamma parameter

lambda - the lambda parameter

Returns:

the baseline

incrementValue

protected static float incrementValue(float baseline,
                                      float step,
                                      float gamma,
                                      float lambda)

Increments a scalar value according to the formula:

v:= v + step - gamma*lambda*v;

Parameters:

baseline - the baseline

step - the step parameter

gamma - the gamma parameter

lambda - the lambda parameter

Returns:

the increment

incrementValue

protected static void incrementValue(org.jblas.FloatMatrix value,
                                     org.jblas.FloatMatrix step,
                                     float gamma,
                                     float lambda)

Increments a vector according to the formula

v:= v + step - gamma*lambda*v

Parameters:

value - the value

step - the step parameter

gamma - the gamma parameter

lambda - the lambda parameter

init

public final void init(Map<String,?> configs,
                       ComputeFunction.InitCallback cb)

Description copied from interface: ComputeFunction

Initialize the ComputeFunction, this is the place to register aggregators.

Specified by:

init in interface ComputeFunction<CfLongId,Svdpp.SvdppValue,Float,FloatMatrixMessage>

Parameters:

configs - configuration parameters

cb - a callback for registering aggregators

masterCompute

public final void masterCompute(int superstep,
                                ComputeFunction.MasterCallback cb)

Description copied from interface: ComputeFunction

A function for performing sequential computations between supersteps.

Specified by:

masterCompute in interface ComputeFunction<CfLongId,Svdpp.SvdppValue,Float,FloatMatrixMessage>

Parameters:

superstep - the superstep

cb - a callback for writing to aggregators or halting the computation

preSuperstep

public void preSuperstep(int superstep,
                         ComputeFunction.Aggregators aggregators)

Description copied from interface: ComputeFunction

Prepare for computation. This method is executed exactly once prior to compute() being called for any of the vertices in the partition.

Specified by:

preSuperstep in interface ComputeFunction<CfLongId,Svdpp.SvdppValue,Float,FloatMatrixMessage>

Parameters:

superstep - the superstep

aggregators - the aggregators

compute

public void compute(int superstep,
                    VertexWithValue<CfLongId,Svdpp.SvdppValue> vertex,
                    Iterable<FloatMatrixMessage> messages,
                    Iterable<EdgeWithValue<CfLongId,Float>> edges,
                    ComputeFunction.Callback<CfLongId,Svdpp.SvdppValue,Float,FloatMatrixMessage> cb)

Description copied from interface: ComputeFunction

The function for computing a new vertex value or sending messages to the next superstep.

Specified by:

compute in interface ComputeFunction<CfLongId,Svdpp.SvdppValue,Float,FloatMatrixMessage>

Parameters:

superstep - the count of the current superstep

vertex - the current vertex with its value

messages - a Map of the source vertex and the message sent from the previous superstep

edges - the adjacent edges with their values

cb - a callback for setting a new vertex value or sending messages to the next superstep

getTotalNumEdges

protected long getTotalNumEdges(ComputeFunction.ReadAggregators aggregators)