Skip to main content

VectorDistance

rig_core::embeddings::distance

Trait VectorDistance 

Source 
pub trait VectorDistance {
    // Required methods
    fn dot_product(&self, other: &Self) -> f64;
    fn cosine_similarity(&self, other: &Self, normalized: bool) -> f64;
    fn euclidean_distance(&self, other: &Self) -> f64;
    fn manhattan_distance(&self, other: &Self) -> f64;
    fn chebyshev_distance(&self, other: &Self) -> f64;

    // Provided method
    fn angular_distance(&self, other: &Self, normalized: bool) -> f64 { ... }
}
Expand description

Similarity and distance metrics of embedding vectors. Supply equal-length vectors; the embedding implementation pairs only their shared prefix. Unnormalized cosine requires nonzero magnitudes.

Required Methods§

Source

fn dot_product(&self, other: &Self) -> f64

The dot product.

Source

fn cosine_similarity(&self, other: &Self, normalized: bool) -> f64

Cosine similarity, from -1 to 1. If normalized is true, the vectors are taken as unit length and the dot product is returned.

Source

fn euclidean_distance(&self, other: &Self) -> f64

The straight-line (L2) distance.

Source

fn manhattan_distance(&self, other: &Self) -> f64

The sum of the coordinates’ absolute differences (L1).

Source

fn chebyshev_distance(&self, other: &Self) -> f64

The largest absolute difference of a coordinate (L-infinity).

Provided Methods§

Source

fn angular_distance(&self, other: &Self, normalized: bool) -> f64

The angle between the vectors as a fraction of pi, from 0 (same direction) to 1 (opposite). normalized is as for Self::cosine_similarity.

Dyn Compatibility§

This trait is not dyn compatible.

In older versions of Rust, dyn compatibility was called "object safety".

Implementors§