Knowledge Commons of Institute of Automation,CAS
An Effective Implicit Multi-interest Interaction Network for Recommendation | |
Yang W(杨威)1; Fan XX(樊鑫鑫)2; Chen YQ(陈逸群)1; Li FM(李非墨)1; Chang HX(常红星)1 | |
2021-12 | |
会议名称 | ICONIP 2021(Neural Information Processing) |
会议日期 | 2021-12-8 |
会议地点 | BALI, Indonesia |
摘要 | Data features in real industrial recommendation scenarios are high-dimensional, diverse and sparse. Rich feature interaction can improve the model effect and bring practical benefits. Factorization machines (FMs) can perform explicit second-order feature interactions, while deep neural networks (DNNs) can perform implicit non-linear feature interactions. A series of models integrating FMs and DNNs are used to perform diverse feature interactions. However, most of the previous work performed feature interaction without considering the diverse interests of users. In reality, users often have multiple preferences and interests, which are implicitly included in the features and need to be effectively extracted. In this paper, we propose an implicit multiple interest network (IMIN), taking into account the importance of interest. Specifically, the model constructs the implicit multiple interests of the user and the item through the implicit multi-interest layer, and realizes the interest alignment between the user and the item through the interest alignment layer. We further use the interest interaction and aggregation layer to construct rich interest feature interactions. In addition, we introduce an auxiliary loss in the model optimization part to ensure the difference of interest. We conducted comprehensive and rich experiments on three real-world data sets. Experimental results show that IMIN performs better than other competitive models, which proves the effectiveness of the model. |
关键词 | Multiple interest Feature interaction Recommendation |
学科门类 | 工学::计算机科学与技术(可授工学、理学学位) |
DOI | 10.1007/978-3-030-92273-3_56 |
URL | 查看原文 |
收录类别 | EI |
语种 | 英语 |
七大方向——子方向分类 | AI芯片与智能计算 |
引用统计 | |
文献类型 | 会议论文 |
条目标识符 | http://ir.ia.ac.cn/handle/173211/47434 |
专题 | 复杂系统认知与决策实验室_高效智能计算与学习 |
通讯作者 | Li FM(李非墨) |
作者单位 | 1.中国科学院自动化研究所 2.北京航空航天大学 |
第一作者单位 | 中国科学院自动化研究所 |
通讯作者单位 | 中国科学院自动化研究所 |
推荐引用方式 GB/T 7714 | Yang W,Fan XX,Chen YQ,et al. An Effective Implicit Multi-interest Interaction Network for Recommendation[C],2021. |
条目包含的文件 | 下载所有文件 | |||||
文件名称/大小 | 文献类型 | 版本类型 | 开放类型 | 使用许可 | ||
会议3 An Effective Imp(227KB) | 会议论文 | 开放获取 | CC BY-NC-SA | 浏览 下载 |
除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。
修改评论