Graph Neural Networks for Social Recommendation

本文探讨了图神经网络在社交推荐系统中的应用,面对用户-项目图的复杂性和社交关系的异质性,提出了一个新颖的模型来解决这些挑战。模型通过迭代聚合特征信息,将用户和项目的潜在因素整合到两个不同的图中,即用户-用户社交图和用户-项目图,以实现更精准的推荐。

Graph Neural Networks for Social Recommendation

1. 摘要

  • 构建基于图神经网络的推荐系统的三大挑战
    • the user-item graph encodes both interactions and their associated opinions
    • social relations have heterogeneous strengths
    • users involve in two graphs (e.g., the user-user social graph and the user-item graph)

2. 介绍

  • 难点

    • Their main idea is how to iteratively aggregate feature information from local graph neighborhoods using neural networks. Meanwhile, node information can be propagated through a graph after transformation and aggregation.
  • GNN 的作用

    • Hence, GNNs naturally integrate the node information as well as the topological structure and have been demonstrated to be powerful in representation learning [ 5 , 7 , 15 ]. On the other hand, data in social recommendation can be represented as graph data with two graphs.

3. 本文模型

model

3.1 用户模型

3.1.1 Item Aggregation

The purpose of item aggregation is to learn item-space user latent factor hiIh_{i}^{I}hiI by considering items a user uiu_{i}ui has interacted with and users’ opinions on these items.

hiI=σ(W⋅Aggreitems(xia,∀a∈C(i))+b)h^{I}_{i} = σ(W · Aggre_{items} ({x_{ia} ,∀a ∈ C(i)}) + b)hiI=σ(WAggreitems(xia,aC(i))+b)

  • hiIh^{I}_{i}hiI: item-space user latent factor

  • C(i)C(i)C(i): item-space user latent factor

  • xiax_{ia}xia: a representation vector to denote opinion-aware interaction between uiu_{i}ui and an item vav_{a}va

The output of MLP is the opinion-aware representation of the interaction between uiu_{i}ui and vav_{a}va,xiax_{ia}xia, as follows:

xia=gv([qa⊕er])x_{ia} = g_{v}([q_{a}⊕e_{r}])xia=gv([qaer])

3.1.2 Social Aggregation

与 Item Aggregation 做法类似

3.2 项目模型

3.2.1 User Aggregation

与 Item Aggregation 做法类似

3.3 预测评分

With the latent factors of users and items (i.e., hih_{i}hi and zjz_{j}zj ), we can first concatenate them [hi⊕zj][h_{i} ⊕ z_{j}][hizj] and then feed it into MLP for rating prediction as:

g1=[hi⊕zj]g_{1} = [h_{i} ⊕ z_{j}]g1=[hizj]

g2=σ(W2⋅g1+b2)g_{2} = σ(W_{2} · g_{1} + b_{2}) g2=σ(W2g1+b2)

gl−1=σ(Wl⋅gl−1+bl)g_{l-1} = σ(W_{l} · g_{l-1} + b_{l}) gl1=σ(Wlgl1+bl)

rij′=wT⋅gl−1r^{′}_{ij} = w^{T} · g_{l−1} rij=wTgl1

  • where l is the index of a hidden layer, and rij′r^{′}_{ij}rij is the predicted rating from uiu_{i}ui to vjv_{j}vj.

3.4 模型训练

Loss function as follows:

Loss=12∣O∣∑i,j∈O(rij′−rij)2Loss = \frac{1}{2|O|} \sum_{i,j∈O} (r^{′}_{ij} − r_{ij})^{2}Loss=2O1i,jO(rijrij)2

  • where ∣O∣|O|O is the number of observed ratings , and rijr_{ij}rij is the ground truth rating assigned by the user i on the item j.

  • Optimizer: RMSprop

  • Overfitting problem: Dropout

4. 实验

4.1 数据集

  • Ciao
  • Epinions

4.2 Baselines

  • PMF
  • SoRec
  • SoReg
  • SocialMF
  • TrustMF
  • NeuMF
  • DeepSoR
  • GCMC+SN

4.3 Result

4.3.1 Performance Comparison of Recommender Systems

1

4.3.2 Model Analysis
  • Effect of Social Network and User Opinions
  • Effect of Attention Mechanisms
  • Effect of Embedding Size

2
3
4

5. 未来工作

  • 探索用户和项目之间的更丰富、复杂的属性
  • 考虑评分和社交关系的动态性
评论
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值