Graph-Based vs Cluster-Based Indexes: HNSW vs IVF for Vector Search
HNSW and IVF are two common ways to make vector search fast without scanning every embedding. HNSW is…
Read moreTopic
Algorithms
ANN search, HNSW, IVF, LSH, and related algorithms.
The Flat Brute-Force Index Explained
A flat, or brute-force, index is the simplest way to perform vector search: store every vector as-is, compare…
Read more
DiskANN: Billion-Scale Search on SSD
DiskANN is a graph-based approximate nearest neighbor search approach designed to make very large vector indexes searchable from…
Read more
Binary Quantization and Hamming Distance in AI Databases
Binary quantization is a way to compress embeddings by turning each numeric dimension into a bit, usually reducing…
Read more
Scalar Quantization Explained
Scalar quantization is a vector compression technique that stores each embedding dimension with fewer bits, most commonly by…
Read more
Product Quantization Explained
Product quantization is a vector compression technique used in large-scale similarity search. It works by splitting each high-dimensional…
Read more
LSH (Locality Sensitive Hashing) Explained
Locality Sensitive Hashing, or LSH, is an approximate nearest neighbor technique that uses special hash functions to place…
Read more
IVF-PQ: Combining Clustering and Compression for Billion-Scale Vector Search
IVF-PQ is a vector search indexing approach that combines inverted file clustering with product quantization so large embedding…
Read more
IVF (Inverted File Index) Explained
IVF, or Inverted File Index, is a vector search indexing method that speeds up nearest-neighbor search by dividing…
Read more
Tuning HNSW: M, efConstruction, and ef
HNSW tuning is about balancing graph quality, search breadth, and resource cost. The M parameter controls how many…
Read more