The Excellence
Directory Platform Template

An efficient clustering method for billion-scale feature vectors that compresses input vectors into short product-quantized (PQ) codes to achieve fast and memory-efficient clustering. PQk-means can cluster one billion 128D SIFT features in 14 hours using just 32 GB of memory.

An advanced vector compression technique for approximate nearest neighbor search that improves upon traditional product quantization by using a hierarchical pyramid structure. Published in 2026, it achieves better compression ratios while maintaining search accuracy.

AHPQ.jl is a Julia library providing training and inference for anisotropic hierarchical product quantization, compatible with ScaNN-style vector quantization and useful for building high-performance vector search pipelines.

Product-Quantization is a GitHub repository implementing the inverted multi-index structure for product-quantization-based approximate nearest neighbor search, providing building blocks for scalable vector search engines.