1、利用不同的行为（click、purchase等）构造不同的user-item graph，从而学出不同graph下的用户和物品表征
2、第一个任务：将上述不同graph下得到的表征融合成一个统一的用户、物品表征，利用BPR进行监督学习
3、第二个任务：利用对比学习构造不同graph下表征彼此之间的关系：
3-1、在目标行为（如purchase）下计算用户之间的相似度，利用相似度选出当前用户positive users以及利用random方法选出negative users
3-2、利用对比学习（设置Loss）使得positive users在不同行为下的表征都能与当前用户在目标行为中的表征相近。
4、联合优化任务一和任务二。
| 文献题目 | 去谷歌学术搜索 | ||||||||||
| Self-supervised Graph Neural Networks for Multi-behavior Recommendation | |||||||||||
| 文献作者 | Shuyun Gu , Xiao Wang , Chuan Shi∗ and Ding Xiao | ||||||||||
| 文献发表年限 | 2022 | ||||||||||
| 文献关键字 | |||||||||||
| GCN; Contrastive learning; Multi-behavior | |||||||||||
| 摘要描述 | |||||||||||
| Traditional recommendation usually focuses on uti- lizing only one target user behavior (e.g., purchase) but ignoring other auxiliary behaviors (e.g., click, add to cart). Early efforts of multi-behavior rec- ommendation often emphasize the differences be- tween multiple behaviors, i.e., they aim to extract useful information by distinguishing different be- haviors. However, the commonality between them, which reflects user’s common preference for items associated with different behaviors, is largely ig- nored. Meanwhile, the multi-behavior recommen- dation still severely suffers from limited supervi- sion signal issue. In this paper, we propose a novel self-supervised graph collaborative filtering model for multi-behavior recommendation named S-MBRec. Specifically, for each behavior, we ex- ecute the GCNs to learn the user and item em- beddings. Then we design a supervised task, dis- tinguishing the importance of different behaviors, to capture the differences between embeddings. Meanwhile, we propose a star-style contrastive learning task to capture the embedding common- ality between target and auxiliary behaviors, so as to alleviate the sparsity of supervision signal, re- duce the redundancy among auxiliary behavior, and extract the most critical information. Finally, we jointly optimize the above two tasks. Extensive experiments, in comparison with state-of-the-arts, well demonstrate the effectiveness of S-MBRec, where the maximum improvement can reach to 20%. | |||||||||||
