Cosine Similarity
Cosine similarity measures the cosine of the angle between two vectors, ranging from -1 to 1, and is the standard similarity metric for embeddings because it is magnitude-invariant.
What is Cosine Similarity?
Cosine similarity measures the cosine of the angle between two vectors, ranging from -1 to 1, and is the standard similarity metric for embeddings because it is magnitude-invariant.
Cosine similarity measures the cosine of the angle between two vectors, ranging from -1 to 1, and is the standard similarity metric for embeddings because it is magnitude-invariant.
Where is it used?
RAG systems retrieve documents by cosine similarity between query and chunk embeddings; recommendation systems and semantic search all rely on it.
How to build it
Use `F.cosine_similarity(emb_a, emb_b, dim=0)` or implement manually as `dot(a,b) / (norm(a) * norm(b))`, and rank documents by similarity to a query embedding.
Code
A practical example: