摘要
图卷积网络如今越来越多地被应用于推荐系统任务中,由于该模型可以有效捕获多跳邻居的信息,因此可以一定程度上缓解数据稀疏性问题,有效提升推荐任务的准确性.但是目前大部分工作都是直接使用图卷积网络,在推荐任务上算法复杂度较高.本文提出了一个融合轻量图卷积网络和注意力机制的模型.该模型通过嵌入传播获得更多邻域的协同信息,同时利用注意力网络对不同的邻域进行区分,最后用于推荐.从而在降低算法复杂度的基础上进一步提升了模型的准确性.通过在Gow alla、Yelp2018和Amazon-book 3个不同领域的真实数据集上的实验结果表明,该方法的性能有较好的表现.
Graph convolution networks are now increasingly used in recommendation system tasks. Because this model can effectively capture the information of multi-hop neighbors,it can alleviate the problem of sparse data to a certain extent and effectively improve the accuracy of recommendation tasks. However,most of the current work is to directly use graph convolution networks,and the algorithm complexity in the recommendation task is relatively high. This paper proposes a model that combines a lightweight graph convolutional network and an attention mechanism. The model obtains more collaborative information of the neighborhood through embedding and propagation,and uses the attention network to distinguish different neighborhoods,and finally used for recommendation.Thus,the accuracy of the model is further improved on the basis of reducing the complexity of the algorithm. The experimental results on the real data sets in three different fields of Gowalla,Yelp2018 and Amazon-book show that the performance of this method has better performance.
作者
郑诚
黄夏炎
ZHENG Cheng;HUANG Xia-yan(School of Computer Science and Technology,Anhui University,Hefei 230601,China;Key Laboratory of Intelligent Computing and Signal Processing,Ministry of Education,Anhui University,Hefei 230601,China)
出处
《小型微型计算机系统》
CSCD
北大核心
2021年第12期2525-2529,共5页
Journal of Chinese Computer Systems
基金
安徽省高校自然科学研究重点项目(KJ2013A020)资助。
关键词
图卷积网络
注意力机制
推荐系统
嵌入传播
graph convolution network
attention mechanism
recommendation system
embedding propagation
