Technique used in large-scale vector quantization for efficient similarity search and data compression by decomposing high-dimensional vectors into smaller sub-vectors and quantizing each sub-vector separately.
Generality: 525
