Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Vector Quantization for Recommender Systems: A Review and Outlook

Qijiong Liu, Xiaoyu Dong, Jiaren Xiao, Nuo Chen, Hengchang Hu, Jieming Zhu, Chenxu Zhu, Tetsuya Sakai, Xiao-Ming Wu

TL;DR

Recommender systems increasingly rely on large representations, making efficiency and scalability critical. The paper surveys vector quantization (VQ) techniques—from standard, parallel, and sequential VQ to differentiable VQ—and organizes VQ4Rec methods into taxonomies by training phase and application scenario, highlighting both efficiency and quality trade-offs. It discusses challenges such as codebook collapse, training with large language models, and multimodal integration, and outlines future directions including item and user tokenization, RS–LLM alignment, and scalable deployment. The findings provide a roadmap for leveraging VQ to enable compact representations, fast retrieval, and effective integration with generative and multimodal recommendation systems.

Abstract

Vector quantization, renowned for its unparalleled feature compression capabilities, has been a prominent topic in signal processing and machine learning research for several decades and remains widely utilized today. With the emergence of large models and generative AI, vector quantization has gained popularity in recommender systems, establishing itself as a preferred solution. This paper starts with a comprehensive review of vector quantization techniques. It then explores systematic taxonomies of vector quantization methods for recommender systems (VQ4Rec), examining their applications from multiple perspectives. Further, it provides a thorough introduction to research efforts in diverse recommendation scenarios, including efficiency-oriented approaches and quality-oriented approaches. Finally, the survey analyzes the remaining challenges and anticipates future trends in VQ4Rec, including the challenges associated with the training of vector quantization, the opportunities presented by large language models, and emerging trends in multimodal recommender systems. We hope this survey can pave the way for future researchers in the recommendation community and accelerate their exploration in this promising field.

Vector Quantization for Recommender Systems: A Review and Outlook

TL;DR

Recommender systems increasingly rely on large representations, making efficiency and scalability critical. The paper surveys vector quantization (VQ) techniques—from standard, parallel, and sequential VQ to differentiable VQ—and organizes VQ4Rec methods into taxonomies by training phase and application scenario, highlighting both efficiency and quality trade-offs. It discusses challenges such as codebook collapse, training with large language models, and multimodal integration, and outlines future directions including item and user tokenization, RS–LLM alignment, and scalable deployment. The findings provide a roadmap for leveraging VQ to enable compact representations, fast retrieval, and effective integration with generative and multimodal recommendation systems.

Abstract

Vector quantization, renowned for its unparalleled feature compression capabilities, has been a prominent topic in signal processing and machine learning research for several decades and remains widely utilized today. With the emergence of large models and generative AI, vector quantization has gained popularity in recommender systems, establishing itself as a preferred solution. This paper starts with a comprehensive review of vector quantization techniques. It then explores systematic taxonomies of vector quantization methods for recommender systems (VQ4Rec), examining their applications from multiple perspectives. Further, it provides a thorough introduction to research efforts in diverse recommendation scenarios, including efficiency-oriented approaches and quality-oriented approaches. Finally, the survey analyzes the remaining challenges and anticipates future trends in VQ4Rec, including the challenges associated with the training of vector quantization, the opportunities presented by large language models, and emerging trends in multimodal recommender systems. We hope this survey can pave the way for future researchers in the recommendation community and accelerate their exploration in this promising field.
Paper Structure (31 sections, 7 equations, 4 figures, 2 tables)

This paper contains 31 sections, 7 equations, 4 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Overview of VQ Techniques
  3. Standard Vector Quantization
  4. Parallel Vector Quantization
  5. Product Quantization (PQ) juang1982multiplepq
  6. Optimized Product Quantization (OPQ) opq
  7. Sequential Vector Quantization
  8. Residual Quantization (RQ) juang1982multiplerq
  9. Additive Quantization (AQ) aq
  10. Differentiable Vector Quantization
  11. Taxonomies of VQ4Rec
  12. Classification by Training Phase
  13. Classification by Application Scenario
  14. Other Classification Frameworks
  15. Efficiency-oriented Approaches
  16. ...and 16 more sections

Figures (4)

  • Figure 1: Interest in VQ4Rec over time. denotes a milestone event or a representative paper.
  • Figure 2: Illustration of the three classical VQ techniques. indicates nearest neighbor search.
  • Figure 3: Integration of VQ techniques with the recommender system at different training stages.
  • Figure 4: Categorization of VQ4Rec methods based on application scenario. The node colors denote different VQ techniques employed. The standard, parallel, and sequential VQ techniques are denoted by green, blue, and red, respectively. The overlap between nodes indicates that the application scenarios they represent share certain similarities.