com.github.jelmerk.knn.scalalike.hnsw
HnswIndex
Related Docs:
class HnswIndex
| package hnsw
Construct a new HnswIndex.
Construct a new HnswIndex.
Type of the external identifier of an item
Type of the vector to perform distance calculation on
Type of items stored in the index
Type of distance between items (expect any numeric type: float, double, int, ..)
dimensionality of the items stored in the index
the distance function
the maximum number of elements in the index
Sets the number of bi-directional links created for every new element during construction. Reasonable range for m is 2-100. Higher m work better on datasets with high intrinsic dimensionality and/or high recall, while low m work better for datasets with low intrinsic dimensionality and/or low recalls. The parameter also determines the algorithm's memory consumption. As an example for d = 4 random vectors optimal m for search is somewhere around 6, while for high dimensional datasets (word embeddings, good face descriptors), higher M are required (e.g. m = 48, 64) for optimal performance at high recall. The range mM = 12-48 is ok for the most of the use cases. When m is changed one has to update the other parameters. Nonetheless, ef and efConstruction parameters can be roughly estimated by assuming that m * efConstruction is a constant.
The size of the dynamic list for the nearest neighbors (used during the search). Higher ef leads to more accurate but slower search. The value ef of can be anything between k and the size of the dataset.
The parameter has the same meaning as ef, but controls the index time / index precision. Bigger efConstruction leads to longer construction, but better index quality. At some point, increasing efConstruction does not improve the quality of the index. One way to check if the selection of ef_construction was ok is to measure a recall for M nearest neighbor search when ef = efConstruction: if the recall is lower than 0.9, then there is room for improvement
enable or disable the experimental remove operation. Indices that support removes will consume more memory
used to serialize the item key during saving of the index. when unspecified java serialization will be used. for the fastest possible save time and smallest indices you will want to provide this
used to serialize the item during saving of the index. when unspecified java serialization will be used. for the fastest possible save time and smallest indices you will want to provide this
used to compare the distances returned by the distance function
the index
Restores a HnswIndex from a File.
Restores a HnswIndex from a File.
Type of the external identifier of an item
Type of the vector to perform distance calculation on
Type of items stored in the index
Type of distance between items (expect any numeric type: float, double, int, ..)
File to read from
the classloader to use
The restored index
Restores a HnswIndex from an InputStream.
Restores a HnswIndex from an InputStream.
Type of the external identifier of an item
Type of the vector to perform distance calculation on
Type of items stored in the index
Type of distance between items (expect any numeric type: float, double, int, ..)
InputStream to restore the index from
the classloader to use
The restored index
Restores a HnswIndex from a Path.
Restores a HnswIndex from a Path.
Type of the external identifier of an item
Type of the vector to perform distance calculation on
Type of items stored in the index
Type of distance between items (expect any numeric type: float, double, int, ..)
Path to read from
The restored index