Learning Binarized Graph Representations with Multi-faceted Quantization Reinforcement for Top-K Recommendation

Yankai Chen; Huifeng Guo; Yingxue Zhang; Chen Ma; Ruiming Tang; Jingjie Li; Irwin King

doi:10.1145/3534678.3539452

Learning Binarized Graph Representations with Multi-faceted Quantization Reinforcement for Top-K Recommendation

Yankai Chen, Huifeng Guo, Yingxue Zhang, Chen Ma, Ruiming Tang, Jingjie Li, Irwin King

Department of Computer Science

Research output: Chapters, Conference Papers, Creative and Literary Works › RGC 32 - Refereed conference paper (with host publication) › peer-review

26 Citations (Scopus)

Abstract

Learning vectorized embeddings is at the core of various recommender systems for user-item matching. To perform efficient online inference, representation quantization, aiming to embed the latent features by a compact sequence of discrete numbers, recently shows the promising potentiality in optimizing both memory and computation overheads. However, existing work merely focuses on numerical quantization whilst ignoring the concomitant information loss issue, which, consequently, leads to conspicuous performance degradation. In this paper, we propose a novel quantization framework to learn Binarized Graph Representations for Top-K Recommendation (BiGeaR). We introduce multi-faceted quantization reinforcement at the pre-, mid-, and post-stage of binarized representation learning, which substantially retains the informativeness against embedding binarization. In addition to saving the memory footprint, it further develops solid online inference acceleration with bitwise operations, providing alternative flexibility for the realistic deployment. The empirical results over five large real-world benchmarks show that BiGeaR achieves about 22%~40% performance improvement over the state-of-the-art quantization-based recommender system, and recovers about 95%~102% of the performance capability of the best full-precision counterpart with over 8× time and space reduction.

Original language	English
Title of host publication	KDD '22: Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining
Publisher	Association for Computing Machinery
Pages	168–178
ISBN (Electronic)	978-1-4503-9385-0
DOIs	https://doi.org/10.1145/3534678.3539452
Publication status	Published - Aug 2022
Event	28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD2022) - Washington DC Convention Center, United States Duration: 14 Aug 2022 → 18 Aug 2022 https://kdd.org/kdd2022/index.html

Publication series

Name	Proceedings of the ACM SIGKDD International Conference on Knowledge Discovery and Data Mining

Conference

Conference	28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD2022)
Country/Territory	United States
Period	14/08/22 → 18/08/22
Internet address	https://kdd.org/kdd2022/index.html

Bibliographical note

Research Unit(s) information for this publication is provided by the author(s) concerned.

Funding

The work described in this paper was partially supported by the National Key Research and Development Program of China (No. 2018AAA0100204), the Research Grants Council of the Hong Kong Special Administrative Region, China (CUHK 2410021, Research Impact Fund, No. R5034-18), and the CUHK Direct Grant (4055147).

Research Keywords

Recommender system
Quantization-based
Embedding Binarization
Graph Convolutional Network
Graph Representation

Access Output

10.1145/3534678.3539452

Other Access Options

Check CityUHK Library

Permanent Link

Right click to copy link

Cite this

Chen, Y., Guo, H., Zhang, Y., Ma, C., Tang, R., Li, J., & King, I. (2022). Learning Binarized Graph Representations with Multi-faceted Quantization Reinforcement for Top-K Recommendation. In KDD '22: Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining (pp. 168–178). (Proceedings of the ACM SIGKDD International Conference on Knowledge Discovery and Data Mining). Association for Computing Machinery. https://doi.org/10.1145/3534678.3539452

Chen, Yankai ; Guo, Huifeng ; Zhang, Yingxue et al. / Learning Binarized Graph Representations with Multi-faceted Quantization Reinforcement for Top-K Recommendation. KDD '22: Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining. Association for Computing Machinery, 2022. pp. 168–178 (Proceedings of the ACM SIGKDD International Conference on Knowledge Discovery and Data Mining).

@inproceedings{6136327654084185bbce33cbb81995b2,

title = "Learning Binarized Graph Representations with Multi-faceted Quantization Reinforcement for Top-K Recommendation",

abstract = "Learning vectorized embeddings is at the core of various recommender systems for user-item matching. To perform efficient online inference, representation quantization, aiming to embed the latent features by a compact sequence of discrete numbers, recently shows the promising potentiality in optimizing both memory and computation overheads. However, existing work merely focuses on numerical quantization whilst ignoring the concomitant information loss issue, which, consequently, leads to conspicuous performance degradation. In this paper, we propose a novel quantization framework to learn Binarized Graph Representations for Top-K Recommendation (BiGeaR). We introduce multi-faceted quantization reinforcement at the pre-, mid-, and post-stage of binarized representation learning, which substantially retains the informativeness against embedding binarization. In addition to saving the memory footprint, it further develops solid online inference acceleration with bitwise operations, providing alternative flexibility for the realistic deployment. The empirical results over five large real-world benchmarks show that BiGeaR achieves about 22%~40% performance improvement over the state-of-the-art quantization-based recommender system, and recovers about 95%~102% of the performance capability of the best full-precision counterpart with over 8× time and space reduction.",

keywords = "Recommender system, Quantization-based, Embedding Binarization, Graph Convolutional Network, Graph Representation",

author = "Yankai Chen and Huifeng Guo and Yingxue Zhang and Chen Ma and Ruiming Tang and Jingjie Li and Irwin King",

note = "Research Unit(s) information for this publication is provided by the author(s) concerned.; 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD2022) ; Conference date: 14-08-2022 Through 18-08-2022",

year = "2022",

month = aug,

doi = "10.1145/3534678.3539452",

language = "English",

series = "Proceedings of the ACM SIGKDD International Conference on Knowledge Discovery and Data Mining",

publisher = "Association for Computing Machinery",

pages = "168–178",

booktitle = "KDD '22: Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining",

address = "United States",

url = "https://kdd.org/kdd2022/index.html",

}

Chen, Y, Guo, H, Zhang, Y, Ma, C, Tang, R, Li, J & King, I 2022, Learning Binarized Graph Representations with Multi-faceted Quantization Reinforcement for Top-K Recommendation. in KDD '22: Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining. Proceedings of the ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, Association for Computing Machinery, pp. 168–178, 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD2022), United States, 14/08/22. https://doi.org/10.1145/3534678.3539452

Learning Binarized Graph Representations with Multi-faceted Quantization Reinforcement for Top-K Recommendation. / Chen, Yankai; Guo, Huifeng; Zhang, Yingxue et al.
KDD '22: Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining. Association for Computing Machinery, 2022. p. 168–178 (Proceedings of the ACM SIGKDD International Conference on Knowledge Discovery and Data Mining).

Research output: Chapters, Conference Papers, Creative and Literary Works › RGC 32 - Refereed conference paper (with host publication) › peer-review

TY - GEN

T1 - Learning Binarized Graph Representations with Multi-faceted Quantization Reinforcement for Top-K Recommendation

AU - Chen, Yankai

AU - Guo, Huifeng

AU - Zhang, Yingxue

AU - Ma, Chen

AU - Tang, Ruiming

AU - Li, Jingjie

AU - King, Irwin

N1 - Research Unit(s) information for this publication is provided by the author(s) concerned.

PY - 2022/8

Y1 - 2022/8

N2 - Learning vectorized embeddings is at the core of various recommender systems for user-item matching. To perform efficient online inference, representation quantization, aiming to embed the latent features by a compact sequence of discrete numbers, recently shows the promising potentiality in optimizing both memory and computation overheads. However, existing work merely focuses on numerical quantization whilst ignoring the concomitant information loss issue, which, consequently, leads to conspicuous performance degradation. In this paper, we propose a novel quantization framework to learn Binarized Graph Representations for Top-K Recommendation (BiGeaR). We introduce multi-faceted quantization reinforcement at the pre-, mid-, and post-stage of binarized representation learning, which substantially retains the informativeness against embedding binarization. In addition to saving the memory footprint, it further develops solid online inference acceleration with bitwise operations, providing alternative flexibility for the realistic deployment. The empirical results over five large real-world benchmarks show that BiGeaR achieves about 22%~40% performance improvement over the state-of-the-art quantization-based recommender system, and recovers about 95%~102% of the performance capability of the best full-precision counterpart with over 8× time and space reduction.

AB - Learning vectorized embeddings is at the core of various recommender systems for user-item matching. To perform efficient online inference, representation quantization, aiming to embed the latent features by a compact sequence of discrete numbers, recently shows the promising potentiality in optimizing both memory and computation overheads. However, existing work merely focuses on numerical quantization whilst ignoring the concomitant information loss issue, which, consequently, leads to conspicuous performance degradation. In this paper, we propose a novel quantization framework to learn Binarized Graph Representations for Top-K Recommendation (BiGeaR). We introduce multi-faceted quantization reinforcement at the pre-, mid-, and post-stage of binarized representation learning, which substantially retains the informativeness against embedding binarization. In addition to saving the memory footprint, it further develops solid online inference acceleration with bitwise operations, providing alternative flexibility for the realistic deployment. The empirical results over five large real-world benchmarks show that BiGeaR achieves about 22%~40% performance improvement over the state-of-the-art quantization-based recommender system, and recovers about 95%~102% of the performance capability of the best full-precision counterpart with over 8× time and space reduction.

KW - Recommender system

KW - Quantization-based

KW - Embedding Binarization

KW - Graph Convolutional Network

KW - Graph Representation

UR - http://www.scopus.com/inward/record.url?scp=85137149085&partnerID=8YFLogxK

UR - https://www.scopus.com/record/pubmetrics.uri?eid=2-s2.0-85137149085&origin=recordpage

U2 - 10.1145/3534678.3539452

DO - 10.1145/3534678.3539452

M3 - RGC 32 - Refereed conference paper (with host publication)

T3 - Proceedings of the ACM SIGKDD International Conference on Knowledge Discovery and Data Mining

SP - 168

EP - 178

BT - KDD '22: Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining

PB - Association for Computing Machinery

T2 - 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD2022)

Y2 - 14 August 2022 through 18 August 2022

ER -

Chen Y, Guo H, Zhang Y, Ma C, Tang R, Li J et al. Learning Binarized Graph Representations with Multi-faceted Quantization Reinforcement for Top-K Recommendation. In KDD '22: Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining. Association for Computing Machinery. 2022. p. 168–178. (Proceedings of the ACM SIGKDD International Conference on Knowledge Discovery and Data Mining). Epub 2022 Aug 14. doi: 10.1145/3534678.3539452

Learning Binarized Graph Representations with Multi-faceted Quantization Reinforcement for Top-K Recommendation

Abstract

Publication series

Conference

Bibliographical note

Funding

Research Keywords

Access Output

Other Access Options

Permanent Link

Other Files and Links

Fingerprint

Cite this