Product Quantization (PQ) search and its derivatives are popular and successful methods for large-scale approximated nearest neighbor search. In this paper, we review the fundamental algorithm of this class of algorithms and provide executable sample codes. We then provide a comprehensive survey of the recent PQ-based methods.
Still struggling with some of the details of product quantizations but this seems like a really good survey... even has a numpy code block implementing PQ.
https://t.co/LOf9gfkS1r https://t.co/QSDIAtZVWk
https://t.co/igROSJWWqS “Fast search with GPUs begins to be studied, which might be one of the next promising directions for ANN research”