Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Use 0.95 as default ratio for lucene radial search traversal similarity #1619

Merged
merged 4 commits into from
Apr 18, 2024
Merged
Show file tree
Hide file tree
Changes from 3 commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 2 additions & 0 deletions src/main/java/org/opensearch/knn/common/KNNConstants.java
Original file line number Diff line number Diff line change
Expand Up @@ -133,4 +133,6 @@ public class KNNConstants {
// Please refer this github issue for more details for choosing this value:
// https://github.com/opensearch-project/k-NN/issues/1049#issuecomment-1694741092
public static int MAX_DISTANCE_COMPUTATIONS = 2048000;

public static float DEFAULT_LUCENE_RADIAL_SEARCH_TRAVERSAL_SIMILARITY_RATIO = 0.95f;
junqiu-lei marked this conversation as resolved.
Show resolved Hide resolved
}
Original file line number Diff line number Diff line change
Expand Up @@ -5,6 +5,7 @@

package org.opensearch.knn.index.query;

import static org.opensearch.knn.common.KNNConstants.DEFAULT_LUCENE_RADIAL_SEARCH_TRAVERSAL_SIMILARITY_RATIO;
import static org.opensearch.knn.common.KNNConstants.VECTOR_DATA_TYPE_FIELD;
import static org.opensearch.knn.index.VectorDataType.SUPPORTED_VECTOR_DATA_TYPES;

Expand Down Expand Up @@ -118,7 +119,13 @@ private static Query getFloatVectorSimilarityQuery(
final float resultSimilarity,
final Query filterQuery
) {
return new FloatVectorSimilarityQuery(fieldName, floatVector, resultSimilarity, filterQuery);
return new FloatVectorSimilarityQuery(
fieldName,
floatVector,
DEFAULT_LUCENE_RADIAL_SEARCH_TRAVERSAL_SIMILARITY_RATIO * resultSimilarity,
resultSimilarity,
filterQuery
);
}

/**
Expand All @@ -131,6 +138,12 @@ private static Query getByteVectorSimilarityQuery(
final float resultSimilarity,
final Query filterQuery
) {
return new ByteVectorSimilarityQuery(fieldName, byteVector, resultSimilarity, filterQuery);
return new ByteVectorSimilarityQuery(
fieldName,
byteVector,
DEFAULT_LUCENE_RADIAL_SEARCH_TRAVERSAL_SIMILARITY_RATIO * resultSimilarity,
resultSimilarity,
filterQuery
);
}
}
Original file line number Diff line number Diff line change
Expand Up @@ -52,6 +52,7 @@
import static org.mockito.Mockito.anyString;
import static org.mockito.Mockito.mock;
import static org.mockito.Mockito.when;
import static org.opensearch.knn.common.KNNConstants.DEFAULT_LUCENE_RADIAL_SEARCH_TRAVERSAL_SIMILARITY_RATIO;
import static org.opensearch.knn.common.KNNConstants.METHOD_HNSW;
import static org.opensearch.knn.index.KNNClusterTestUtils.mockClusterService;
import static org.opensearch.knn.index.util.KNNEngine.ENGINES_SUPPORTING_RADIAL_SEARCH;
Expand Down Expand Up @@ -394,7 +395,12 @@ public void testDoToQuery_whenNormal_whenDoRadiusSearch_whenDistanceThreshold_th
KNNMethodContext knnMethodContext = new KNNMethodContext(KNNEngine.LUCENE, SpaceType.L2, methodComponentContext);
when(mockKNNVectorField.getKnnMethodContext()).thenReturn(knnMethodContext);
FloatVectorSimilarityQuery query = (FloatVectorSimilarityQuery) knnQueryBuilder.doToQuery(mockQueryShardContext);
assertTrue(query.toString().contains("resultSimilarity=" + KNNEngine.LUCENE.distanceToRadialThreshold(MAX_DISTANCE, SpaceType.L2)));
float resultSimilarity = KNNEngine.LUCENE.distanceToRadialThreshold(MAX_DISTANCE, SpaceType.L2);

assertTrue(query.toString().contains("resultSimilarity=" + resultSimilarity));
assertTrue(
query.toString().contains("traversalSimilarity=" + DEFAULT_LUCENE_RADIAL_SEARCH_TRAVERSAL_SIMILARITY_RATIO * resultSimilarity)
);
}

public void testDoToQuery_whenNormal_whenDoRadiusSearch_whenScoreThreshold_thenSucceed() {
Expand Down
Loading