readme.org

Spherical Graph Embedding for Item Retrieval in Recommendation System

Source code for CIKM’22 paper(sp) “Spherical Graph Embedding for Item Retrieval in Recommendation System”. (link: https://dl.acm.org/doi/10.1145/3511808.3557704)

Introduction

In this paper, we present a simple and effective graph-based recommendation method. Internally, we call it Q-align, which has widely adopted in many different scenarios:

Item/User Retrieval

Graph Transfer Learning

Diversity in recommender

Universal Graph Embedding

etc

There are 20+ launches at Bytedance. The wide deployments of the proposed method show its effectiveness in real-life Recommendation Systems.

Thanks

To evaluate our method on the public dataset, we use the CogDL Toolkit(https://cogdl.ai/). The results are listed in “result.txt”. We’re grateful to the authors of the CogDL Toolkit(https://cogdl.ai) for their excellent work, which saves us much time.

Citation

@inproceedings{10.1145/3511808.3557704,
author = {Zhu, Wenqiao and Xu, Yesheng and Huang, Xin and Min, Qiyang and Zhou, Xun},
title = {Spherical Graph Embedding for Item Retrieval in Recommendation System},
year = {2022},
pages = {4752–4756},
series = {CIKM '22}
}