Alternative ANN techniques and families

Alternative ANN techniques and families place HNSW among hashing, trees, inverted files, product-quantization indexes, and hybrid compositions - so you can choose an index family on workload grounds, not fashion.
Created: Updated: 2 min read

These are the topics in this section, each on its own page with a stable path you can bookmark or share.

What topics are covered in this section?

Alternative ANN techniques and families place HNSW among hashing, trees, inverted files, product-quantization indexes, and hybrid compositions – so you can choose an index family on purpose.

What else exists besides proximity graphs?

Locality-sensitive hashing and hashing-based ANN probe buckets. Inverted file indexes, coarse quantizers, and cluster-based candidate reduction open posting lists near the query. Tree-based methods – k-d, ball, R, VP trees – partition space recursively and struggle more as dimension grows. Product-quantization indexes compress lists for memory. Hybrid and composite or two-level structures combine families. Exact brute-force baselines keep everyone honest. The taxonomy and trade-off pages help you match method to N, d, and update patterns.

HNSW is often best for high-recall interactive search on embeddings, not the only tool.

How does Weaviate sit in this taxonomy?

Weaviate offers HNSW, flat, dynamic (flat then HNSW), and HFresh indexes – spanning exact small-set search, graph ANN, and disk-oriented cluster search. Knowing IVF and LSH vocabulary still helps when reading papers or migrating workloads. Choose flat for tiny tenants, HNSW for large in-memory QPS, HFresh when RAM dominates cost.

Part V and Part X compare compositions and graph alternatives in more depth.

What reading path helps?

Start with the taxonomy, IVF, LSH, and tree-based ANN pages, then hybrid indexes. Continue into graph-based alternatives for NSW, NSG, Vamana, and DiskANN-style designs.

This section maps the ANN landscape around HNSW. Next, read “What is the taxonomy of ANN methods?” or “What is an inverted file index (IVF)?”, then pick Weaviate index types with that map in mind.