Scope & Purpose

Add inner product metric for vector index. Inner product metric is similar to the cosine metric except we won't normalize vectors on search and insertion. The example of using the inner product is here:

  db.<collection>.ensureIndex({
    type: "vector",
    fields: ["value"],
    params: {
      "metric": "innerProduct",
      "dimension": 300,
      "nLists": 100
    }
  });

Also adding a new AQL function APPROX_NEAR_INNER_PRODUCT that uses vector index built with inner product metric. And using it in the query looks like this:

FOR d IN col
SORT APPROX_NEAR_INNER_PRODUCT(d.vector, @qp) DESC
LIMIT 3
RETURN d; 

Like in cosine metric with inner product metric, the bigger the similarity score more similar the vectors are, meaning we need to use DESC to get similar vectors.

• 💩 Bugfix
• 🍕 New feature
• 🔥 Performance improvement
• 🔨 Refactoring/simplification

Checklist

• Tests
  • Regression tests
  • C++ Unit tests
  • integration tests
  • resilience tests
• 📖 CHANGELOG entry made
• 📚 documentation written (release notes, API changes, ...)
• Backports
  • Backport for 3.12.0: (Please link PR)
  • Backport for 3.11: (Please link PR)
  • Backport for 3.10: (Please link PR)

Related Information

(Please reference tickets / specification / other PRs etc)

• Docs PR:
• Enterprise PR:
• GitHub issue / Jira ticket:
• Design document: