Struct PQCodebook 

pub struct PQCodebook {
    pub num_subvectors: usize,
    pub num_centroids: usize,
    pub sub_dim: usize,
    pub centroids: Vec<Vec<Vec<f32>>>,
}

Fields

num_subvectors: usize

Number of sub-vectors (M)

num_centroids: usize

Number of centroids per sub-space (K, typically 256 so codes fit in u8)

sub_dim: usize

Dimensionality of each sub-vector = dim / num_subvectors

centroids: Vec<Vec<Vec<f32>>>

Centroids: [num_subvectors][num_centroids][sub_dim]

Implementations

impl PQCodebook

pub fn train( vectors: &[Vec<f32>], num_subvectors: usize, num_centroids: usize, max_iter: usize, ) -> Result<Self, AilakeError>

Train PQ codebook via k-means on each sub-space independently.

pub fn train_with_kmeans<F>( vectors: &[Vec<f32>], num_subvectors: usize, num_centroids: usize, max_iter: usize, kmeans_fn: F, ) -> Result<Self, AilakeError>

where F: Fn(&[Vec<f32>], usize, usize) -> Vec<Vec<f32>>,

Train PQ codebook with a custom k-means backend (e.g. GPU-accelerated).

kmeans_fn(vecs, k, max_iter) must return exactly k centroids of the same dimensionality as vecs. The built-in CPU path passes kmeans.

pub fn encode(&self, vector: &[f32]) -> Vec<u8>

Encode a single vector into num_subvectors u8 codes.

pub fn decode(&self, codes: &[u8]) -> Vec<f32>

Decode codes back into an approximate vector (centroid reconstruction).

pub fn compute_adc_table(&self, query: &[f32]) -> Vec<Vec<f32>>

Precompute query-to-centroid L2 distances for Asymmetric Distance Computation. Returns [num_subvectors][num_centroids] distance table. ADC is O(M*K) per query precomputation, then O(M) per encoded vector — much faster than symmetric distance which would require decoding each vector first.

pub fn adc_distance(&self, codes: &[u8], table: &[Vec<f32>]) -> f32

Compute approximate L2 distance using the precomputed ADC table.

Trait Implementations

impl Clone for PQCodebook

fn clone(&self) -> PQCodebook

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for PQCodebook

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for PQCodebook

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for PQCodebook

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.