Package org.apache.lucene.util.quantization

Class QuantizedByteVectorValues

java.lang.Object

org.apache.lucene.index.KnnVectorValues

org.apache.lucene.index.ByteVectorValues

org.apache.lucene.util.quantization.BaseQuantizedByteVectorValues

org.apache.lucene.util.quantization.QuantizedByteVectorValues

All Implemented Interfaces:

HasIndexSlice

Direct Known Subclasses:

OffHeapScalarQuantizedVectorValues

public abstract class QuantizedByteVectorValues
extends BaseQuantizedByteVectorValues

Scalar quantized byte vector values

WARNING: This API is experimental and might change in incompatible ways in the next release.

Nested Class Summary

Nested Classes

Modifier and Type	Class	Description
static enum	QuantizedByteVectorValues.ScalarEncoding	Allowed encodings for scalar quantization.

Nested classes/interfaces inherited from class org.apache.lucene.index.KnnVectorValues

KnnVectorValues.DocIndexIterator

Constructor Summary

Constructors

Constructor	Description
QuantizedByteVectorValues()

Method Summary

Modifier and Type	Method	Description
abstract QuantizedByteVectorValues	copy()	Creates a new copy of this KnnVectorValues.
abstract float[]	getCentroid()
abstract float	getCentroidDP()
abstract OptimizedScalarQuantizer.QuantizationResult	getCorrectiveTerms(int vectorOrd)	Retrieve the corrective terms for the given vector ordinal.
abstract OptimizedScalarQuantizer	getQuantizer()
abstract QuantizedByteVectorValues.ScalarEncoding	getScalarEncoding()

Methods inherited from class org.apache.lucene.util.quantization.BaseQuantizedByteVectorValues

getSlice, scorer

Methods inherited from class org.apache.lucene.index.ByteVectorValues

checkField, fromBytes, getEncoding, rescorer, scorer, vectorValue

Methods inherited from class org.apache.lucene.index.KnnVectorValues

createDenseIterator, createSparseIterator, dimension, fromDISI, getAcceptOrds, getVectorByteLength, iterator, ordToDoc, prefetch, size

Methods inherited from class java.lang.Object

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

Constructor Details

QuantizedByteVectorValues

public QuantizedByteVectorValues()

Method Details

getCorrectiveTerms

public abstract OptimizedScalarQuantizer.QuantizationResult getCorrectiveTerms(int vectorOrd)
                                                                        throws IOException

Retrieve the corrective terms for the given vector ordinal. For the dot-product family of distances, the corrective terms are, in order

• the lower optimized interval
• the upper optimized interval
• the dot-product of the non-centered vector with the centroid
• the sum of quantized components

For euclidean:

• the lower optimized interval
• the upper optimized interval
• the l2norm of the centered vector
• the sum of quantized components

Parameters:

vectorOrd - the vector ordinal

Returns:

the corrective terms

Throws:

IOException - if an I/O error occurs

getQuantizer

public abstract OptimizedScalarQuantizer getQuantizer()

Returns:

the quantizer used to quantize the vectors

getScalarEncoding

public abstract QuantizedByteVectorValues.ScalarEncoding getScalarEncoding()

Returns:

the scalar encoding used to pack the stored vectors.

getCentroid

public abstract float[] getCentroid()
                             throws IOException

Returns:

the centroid used to center the vectors prior to quantization

Throws:

IOException

getCentroidDP

public abstract float getCentroidDP()
                             throws IOException

Returns:

the dot product of the centroid.

Throws:

IOException

copy

public abstract QuantizedByteVectorValues copy()
                                        throws IOException

Description copied from class: KnnVectorValues

Creates a new copy of this KnnVectorValues. This is helpful when you need to access different values at once, to avoid overwriting the underlying vector returned.

Specified by:

copy in class ByteVectorValues

Throws:

IOException