Generalized residual vector quantization for large scale data

TL;DR

This paper introduces a generalized residual vector quantization framework that iteratively reduces quantization error, significantly improving accuracy and efficiency for large scale data tasks like search and retrieval.

Contribution

The paper proposes a novel generalized residual vector quantization (GRVQ) framework that unifies and improves upon existing residual vector quantization methods.

Findings

GRVQ outperforms existing methods in quantization accuracy

GRVQ is more computationally efficient

GRVQ effectively handles large scale datasets

Abstract

Vector quantization is an essential tool for tasks involving large scale data, for example, large scale similarity search, which is crucial for content-based information retrieval and analysis. In this paper, we propose a novel vector quantization framework that iteratively minimizes quantization error. First, we provide a detailed review on a relevant vector quantization method named \textit{residual vector quantization} (RVQ). Next, we propose \textit{generalized residual vector quantization} (GRVQ) to further improve over RVQ. Many vector quantization methods can be viewed as the special cases of our proposed framework. We evaluate GRVQ on several large scale benchmark datasets for large scale search, classification and object retrieval. We compared GRVQ with existing methods in detail. Extensive experiments demonstrate our GRVQ framework substantially outperforms existing methods in term of quantization accuracy and computation efficiency.