Optimized Codebook Construction and Assignment for Product Quantization-based Approximate Nearest Neighbor Search

Nearest neighbor search (NNS) among large-scale and high-dimensional vectors has played an important role in recent large-scale multimedia search applications. This paper proposes an optimized codebook construction algorithm for approximate NNS based on product quantization. The proposed algorithm iteratively optimizes both codebooks for product quantization and an assignment table that indicates the optimal codebook in product quantization. In experiments, the proposed method is shown to achieve better accuracy in approximate NNS than the conventional method with the same memory requirement and the same computational cost. Furthermore, use of a larger number of codebooks increases the accuracy of approximate NNS at the expense of a slight increase in the memory requirement.