Spectre.Classifier.Math (Spectre v0.3.0)

Small vector math helpers used by classifier and embedding strategies.

These functions stay pure and dependency-free so classifier scoring can be tested without loading embedding models or Vettore indexes.

Summary

Returns a normalized centroid for vectors.

Returns cosine similarity in [-1.0, 1.0].

Normalizes a vector to unit length.

Returns Vettore's raw cosine score.

@spec centroid([[number()]]) :: [float()]

Returns a normalized centroid for vectors.

@spec cosine([number()], [number()]) :: float()

Returns cosine similarity in [-1.0, 1.0].

@spec normalize([number()]) :: [float()]

Normalizes a vector to unit length.

@spec raw_cosine_score(term()) :: float()

Returns Vettore's raw cosine score.