org.lenskit.mf.funksvd

Class FunkSVDUpdater

java.lang.Object

org.lenskit.mf.funksvd.FunkSVDUpdater

public final class FunkSVDUpdater
extends Object

Encapsulation of the FunkSVD update process. Using this class takes a two-step process:

1. Call prepare(int, double, double, double, double, double) to prepare an update.
2. Call getItemFeatureUpdate() and getUserFeatureUpdate() to get the deltas to apply to the item-feature and user-feature values, respectively.

The updater can be reused for multiple updates, but cannot be shared between threads. It is typical to create one updater and reuse it for much of the training process.

The updater also tracks statistics across runs.

Since:

2.1

Method Summary

Modifier and Type	Method and Description
double	getError() Get the error from the prepared update.
double	getItemFeatureUpdate() Get the update for the item-feature value.
double	getRMSE() Get the RMSE of all updates done since the last reset.
int	getUpdateCount() Get the number of updates this updater has prepared since the last reset.
double	getUserFeatureUpdate() Get the update for the user-feature value.
void	prepare(int feature, double rating, double estimate, double uv, double iv, double trail) Prepare the updater for updating the feature values for a particular user/item ID.
void	resetStatistics() Reset the statistics and counters tracked by this updater.

Methods inherited from class java.lang.Object

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

Method Detail

resetStatistics

public void resetStatistics()

Reset the statistics and counters tracked by this updater.

getUpdateCount

public int getUpdateCount()

Get the number of updates this updater has prepared since the last reset.

Returns:

The number of updates done.

See Also:

resetStatistics()

getRMSE

public double getRMSE()

Get the RMSE of all updates done since the last reset.

Returns:

The root-mean-squared error of the updates since the last reset.

prepare

public void prepare(int feature,
                    double rating,
                    double estimate,
                    double uv,
                    double iv,
                    double trail)

Prepare the updater for updating the feature values for a particular user/item ID.

Parameters:

feature - The feature we are training.

rating - The rating value.

estimate - The estimate through the previous feature.

uv - The user feature value.

iv - The item feature value.

trail - The sum of the trailing feature value products.

getError

public double getError()

Get the error from the prepared update.

Returns:

The estimation error in the prepared update.

getUserFeatureUpdate

public double getUserFeatureUpdate()

Get the update for the user-feature value.

Returns:

The delta to apply to the user-feature value.

getItemFeatureUpdate

public double getItemFeatureUpdate()

Get the update for the item-feature value.

Returns:

The delta to apply to the item-feature value.