Package org.apache.lucene.util.quantization

Interface ScalarQuantizedVectorSimilarity

All Known Implementing Classes:

ScalarQuantizedVectorSimilarity.DotProduct, ScalarQuantizedVectorSimilarity.Euclidean, ScalarQuantizedVectorSimilarity.MaximumInnerProduct

public interface ScalarQuantizedVectorSimilarity

Calculates and adjust the scores correctly for quantized vectors given the scalar quantization parameters

Nested Class Summary

Nested Classes

Modifier and Type	Interface	Description
static interface	ScalarQuantizedVectorSimilarity.ByteVectorComparator	Compares two byte vectors
static class	ScalarQuantizedVectorSimilarity.DotProduct	Calculates dot product on quantized vectors, applying the appropriate corrections
static class	ScalarQuantizedVectorSimilarity.Euclidean	Calculates euclidean distance on quantized vectors, applying the appropriate corrections
static class	ScalarQuantizedVectorSimilarity.MaximumInnerProduct	Calculates max inner product on quantized vectors, applying the appropriate corrections

Method Summary

Modifier and Type	Method	Description
static ScalarQuantizedVectorSimilarity	fromVectorSimilarity(VectorSimilarityFunction sim, float constMultiplier, byte bits)	Creates a ScalarQuantizedVectorSimilarity from a VectorSimilarityFunction and the constant multiplier used for quantization.
float	score(byte[] queryVector, float queryVectorOffset, byte[] storedVector, float vectorOffset)

Method Detail

fromVectorSimilarity

static ScalarQuantizedVectorSimilarity fromVectorSimilarity(VectorSimilarityFunction sim,
                                                            float constMultiplier,
                                                            byte bits)

Creates a ScalarQuantizedVectorSimilarity from a VectorSimilarityFunction and the constant multiplier used for quantization.

Parameters:

sim - similarity function

constMultiplier - constant multiplier used for quantization

bits - number of bits used for quantization

Returns:

a ScalarQuantizedVectorSimilarity that applies the appropriate corrections

score

float score(byte[] queryVector,
            float queryVectorOffset,
            byte[] storedVector,
            float vectorOffset)