Crate ann_dataset

A lightweight research library for managing Approximate Nearest Neighbor search datasets.

It offers the following features:

• Storage of dense, sparse, and dense-sparse vector sets;
• Storage of query sets with ground-truth (i.e., exact nearest neighbors) according to different metrics;
• Basic functionality such as computing recall given a retrieved set; and,
• Serialization into and deserialization from HDF5 file format.

Structs

• GroundTruth
  Defines the exact nearest neighbors.
• InMemoryAnnDataset
  An ANN dataset.
• PointSet
  A set of points (dense, sparse, or both) represented as a matrix, where each row corresponds to a single vector.
• QuerySet
  A set of query points (dense, sparse, or both) and their exact nearest neighbors for various metrics.

Enums

• Metric
  Collection of metrics and distance functions that characterize an ANN search.

Traits

• AnnDataset
• Hdf5File
• Hdf5Serialization