Product Quantization (PQ)

How product quantization is used in image retrieval:

"' After quantitative learning, for a given query sample, the query sample and library can be calculated by looking up a tableAsymmetric distance of the samples in"'

A brief description of product quantization: The typical representative of vector quantization methods is the product quantization (PQ, Product
Quantization) method, which decomposes the feature space into Cartesian products of multiple low-dimensional subspaces, and then quantize each subspace individually.In the training phase, each subspace is clustered to obtain kk centroids (ie quantizers), and the Cartesian product of all these centroids constitutes a dense division of the whole space, and can ensure that the quantization error is relatively small;After quantitative learning, for a given query sample, the asymmetric distance between the query sample and the sample in the library can be calculated by looking up the table.Approximate Nearest Neighbor Search