org.grouplens.lenskit.transform.normalize

Interface VectorNormalizer

All Known Implementing Classes:

AbstractVectorNormalizer, IdentityVectorNormalizer, MeanCenteringVectorNormalizer, MeanVarianceNormalizer, UnitVectorNormalizer

@DefaultImplementation(value=IdentityVectorNormalizer.class)
public interface VectorNormalizer

Normalize vectors by applying a reversible transformation with respect to a reference vector. The reference vector is used to compute the normalization, and it is applied to the target vector; this allows e.g. the user's average rating to be subtracted from a set of ratings.

Method Summary

Modifier and Type	Method and Description
VectorTransformation	makeTransformation(SparseVector reference) Create a vector transformation that normalizes and denormalizes vectors with respect to the specified entity.
MutableSparseVector	normalize(SparseVector reference, MutableSparseVector target) Normalize a vector in-place with a reference vector.

Method Detail

normalize

MutableSparseVector normalize(@Nonnull
                              SparseVector reference,
                              @Nullable
                              MutableSparseVector target)

Normalize a vector in-place with a reference vector.

To understand the relationship of reference and vector, consider wanting to subtract the user's mean rating from a set of ratings. To do that, the user's rating vector is reference, and the vector of ratings to be adjusted is vector.

This method is equivalent to makeTransformation(reference).apply(target).

Parameters:

reference - The reference used to compute whatever transformation is needed (e.g. the mean value).

target - The vector to normalize. If null, a new mutable copy of reference is created.

Returns:

target, or a normalized mutable copy of reference if target is null.

makeTransformation

VectorTransformation makeTransformation(SparseVector reference)

Create a vector transformation that normalizes and denormalizes vectors with respect to the specified entity. This allows transformations to be applied multiple times to different vectors and also unapplied.

If the reference vector is empty, the returned transformation should be the identity transform. Results are undefined if the reference vector is not complete or contains NaN values.

If the normalization needs to retain a copy of the sparse vector, it will take an immutable copy.

Parameters:

reference - The reference vector.

Returns:

A transformation built from the reference vector.