Mnemosyne.Graph.Similarity (mnemosyne v0.1.5)

Cosine similarity computations using Scholar and Nx.

Provides vector similarity scoring for knowledge graph node retrieval.

Summary

Computes cosine similarity between two vectors, returning a value in [-1, 1].

Returns the k most similar candidates to query, sorted by descending similarity.

@spec cosine_similarity([float()], [float()]) :: float()

Computes cosine similarity between two vectors, returning a value in [-1, 1].

@spec top_k([float()], [{String.t(), [float()] | nil}], non_neg_integer()) :: [
  {String.t(), float()}
]

Returns the k most similar candidates to query, sorted by descending similarity.