Package org.apache.lucene.search.knn

Class KnnSearchStrategy.Hnsw

java.lang.Object
org.apache.lucene.search.knn.KnnSearchStrategy
org.apache.lucene.search.knn.KnnSearchStrategy.Hnsw
Direct Known Subclasses:
KnnSearchStrategy.Patience
Enclosing class:
KnnSearchStrategy
public static class KnnSearchStrategy.Hnsw extends KnnSearchStrategy
A strategy for kNN search that uses HNSW
WARNING: This API is experimental and might change in incompatible ways in the next release.

  • Field Details

  • Constructor Details

    • Hnsw

      public Hnsw(int filteredSearchThreshold)
      Create a new Hnsw strategy
      Parameters:
      filteredSearchThreshold - threshold for filtered search, a percentage value from 0 to 100 where 0 means never use filtered search and 100 means always use filtered search.

  • Method Details

    • filteredSearchThreshold

      public int filteredSearchThreshold()

    • useFilteredSearch

      public final boolean useFilteredSearch(float ratioPassingFilter)
      Whether to use filtered search based on the ratio of vectors that pass the filter
      Parameters:
      ratioPassingFilter - ratio of vectors that pass the filter
      Returns:
      true if filtered search should be used

    • equals

      public boolean equals(Object o)
      Description copied from class: KnnSearchStrategy
      Override and implement search strategy instance equivalence properly in a subclass.
      Specified by:
      equals in class KnnSearchStrategy

    • hashCode

      public int hashCode()
      Description copied from class: KnnSearchStrategy
      Override and implement search strategy hash code properly in a subclass.
      Specified by:
      hashCode in class KnnSearchStrategy
      See Also:

    • nextVectorsBlock

      public void nextVectorsBlock()
      Description copied from class: KnnSearchStrategy
      Signal processing of the next block of vectors.
      Specified by:
      nextVectorsBlock in class KnnSearchStrategy