What it is: A graph-based algorithm that organizes vectors in multiple layers, like a highway system for finding similar vectors.
How it works:
- Creates a multi-level graph where each vector connects to its nearest neighbors
- Top layers have fewer nodes but longer connections (like highways)
- Bottom layers have more nodes with local connections (like city streets)
- Search starts at the top layer and navigates down to find the closest match
Why it matters: Very fast and accurate for similarity search, especially with millions of vectors. Used by many modern vector databases.
Real-world analogy: Like using Google Maps - you start with major highways to get to the right area, then use local streets to find the exact address.