Cosine Similarity

Definition

The most common mathematical metric used to measure how similar two vectors are. It measures the cosine of the angle between two multi-dimensional vectors.

Explain Like I'm New

Imagine two arrows shooting out from the center of a graph. If the arrows point in the exact same direction, they are identical concepts (Score = 1). If they form a 90-degree angle, they are completely unrelated (Score = 0). If they point in exact opposite directions, they are exact opposites (Score = -1).

Real World Example

In a Vector Database, when you ask a question, the database calculates the Cosine Similarity between your Question Vector and every Document Vector. It sorts the list, and the document with a score closest to 1.0 is returned as the best answer.

Common Use Cases

•Vector comparisons
•Similarity scoring

Interview Questions

basic

If two vectors have a Cosine Similarity score of 0.99, are they highly similar or highly different?

intermediate

Why is Cosine Similarity generally preferred over 'Euclidean Distance' for text embeddings?

Flash Cards

Question

Similar or different?

Click to reveal answer

Answer

Highly similar. A score of 1.0 is a perfect match.

Question

Why Cosine over Euclidean?

Click to reveal answer

Answer

Euclidean Distance measures the physical distance between the tips of the arrows. If a document mentions 'Apple' 5 times, and another mentions it 500 times, their Euclidean distance is huge. Cosine measures the *angle*. Since both documents point in the 'Apple' direction, Cosine recognizes they are about the same topic, regardless of document length.