HNSW (Hierarchical Navigable Small World)
Multi-Layer Proximity Graphs for Sub-Millisecond Approximate Nearest Neighbor Search
#ANN#Vector Search#HNSW#Graphs#Skip-List#Vector DB
Choose Presentation Mode:
STAGE 1 / 7— Anti-Pattern
Section 1: The LLM Anti-Pattern vs Right-Sized Model
The Naive Generative LLM Approach:
Writing a linear brute-force scan or prompt to locate the nearest vector among 10 million 1536-dimensional embeddings.
Why It Fails in Production:
Brute-force scan takes seconds and wastes gigabytes of memory; HNSW locates the top nearest neighbors in 1ms with 99% recall.
Targeted Algorithm (HNSW (Hierarchical Navigable Small World))
Latency:1ms (HNSW)
Cost / 1M Ops:$0.00
Determinism:>98% Recall-at-10
Generative LLM Alternative
Latency:N/A (Impossible)
Cost / 1M Ops:N/A
Determinism:Fails to Index