Graph Neural Networks for Social Recommendation

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

3.1 用户模型

3.1.1 Item Aggregation

The purpose of item aggregation is to learn item-space user latent factor $h_i^I$ by considering items a user $u_i$ has interacted with and users’ opinions on these items.

$$h_i^I=\sigma (W\cdot Aggr{e}_{items}(x_{ia},\forall a\in C(i))+b)$$

• $h_i^I$: item-space user latent factor

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

• $x_{ia}$: a representation vector to denote opinion-aware interaction between $u_i$ and an item $v_a$

The output of MLP is the opinion-aware representation of the interaction between $u_i$ and $v_a$,$x_{ia}$, as follows:

$$x_{ia}=g_v([q_a\oplus e_r])$$

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., $h_i$ and $z_j$ ), we can first concatenate them $[h_i\oplus z_j]$ and then feed it into MLP for rating prediction as:

$$g_1=[h_i\oplus z_j]$$

$$g_2=\sigma (W_2\cdot g_1+b_2)$$

$$g_{l-1}=\sigma (W_l\cdot g_{l-1}+b_l)$$

$$r_{ij}^{\prime }=w^T\cdot g_{l-1}$$

• where l is the index of a hidden layer, and $r_{ij}^{\prime }$ is the predicted rating from $u_i$ to $v_j$.

3.4 模型训练

Loss function as follows:

$$Loss=\frac{1}{2|O|}\sum _{i,j\in O}(r_{ij}^{\prime }-r_{ij}{)}^2$$

• where $|O|$ is the number of observed ratings , and $r_{ij}$ 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

4.3.2 Model Analysis

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

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