Stacked Quantizers for Compositional Vector Compression

TL;DR

This paper introduces Stacked Quantizers, a hierarchical vector compression method that balances the accuracy of Additive Quantization with the speed of Product Quantization, suitable for high-dimensional feature datasets.

Contribution

It proposes a novel hierarchical quantization approach that improves compression accuracy while significantly reducing encoding complexity compared to existing methods.

Findings

Achieves quantization error comparable or lower than Additive Quantization

Offers several orders of magnitude faster encoding than AQ

Performs well on standard and neural network feature datasets

Abstract

Recently, Babenko and Lempitsky introduced Additive Quantization (AQ), a generalization of Product Quantization (PQ) where a non-independent set of codebooks is used to compress vectors into small binary codes. Unfortunately, under this scheme encoding cannot be done independently in each codebook, and optimal encoding is an NP-hard problem. In this paper, we observe that PQ and AQ are both compositional quantizers that lie on the extremes of the codebook dependence-independence assumption, and explore an intermediate approach that exploits a hierarchical structure in the codebooks. This results in a method that achieves quantization error on par with or lower than AQ, while being several orders of magnitude faster. We perform a complexity analysis of PQ, AQ and our method, and evaluate our approach on standard benchmarks of SIFT and GIST descriptors, as well as on new datasets of features obtained from state-of-the-art convolutional neural networks.