Struct FlatIvfIndex 

pub struct FlatIvfIndex {
    pub dim: usize,
    pub num_cells: usize,
    pub cells: Vec<IvfCell>,
    pub centroids: Vec<Centroid>,
}

Flat IVF index.

Fields

dim: usize

num_cells: usize

cells: Vec<IvfCell>

centroids: Vec<Centroid>

Implementations

impl FlatIvfIndex

pub fn new(dim: usize, num_cells: usize) -> Self

Create an untrained index with num_cells cells and vectors of dimension dim.

pub fn train(&mut self, vectors: &[Vec<f32>])

Train the index using Lloyd’s k-means (up to 20 iterations) on vectors.

Centroids are initialised by selecting evenly-spaced samples from the input.

pub fn insert(&mut self, id: u64, vector: Vec<f32>)

Insert a vector with the given id into the nearest cell.

pub fn search( &self, query: &[f32], k: usize, n_probe: usize, ) -> Vec<SearchResult>

Search for the k nearest neighbours of query, probing n_probe cells.

pub fn remove(&mut self, id: u64) -> bool

Remove the vector with id from the index. Returns true if found.

pub fn len(&self) -> usize

Total number of vectors in the index.

pub fn is_empty(&self) -> bool

Returns true if no vectors are stored.

pub fn nearest_centroid(&self, vec: &[f32]) -> usize

Return the index of the nearest cell for vec.

pub fn l2_distance(a: &[f32], b: &[f32]) -> f32

Compute the squared L2 distance between two equal-length slices.