异质图神经网络相关论文（2023年）

异质图神经网络，指图中节点的类型大于1或者边的类型大于1。近期做异质图神经网络的，我知道的有：1）meta-path，人为定义的有语义信息的路径，比如 电影—导演—电影 这样一个元路径，电影和导演是不同的节点类型，这个路径的语义信息就是这两部电影是由同一个导演做的， 基于元路径再去做图神经网络的节点更新等等。

2）第二种是对第一种的改进，是说人为定义，预先设置好的元路径可能捕捉不到一些语义信息，提出了一种可以学习的元路径。

3）对不同的边，使用不同的邻接矩阵，就是有多个A，分别做图卷积，再拼接之类的操作，比如用两个邻接矩阵分别表示有向边的出边和入边。

应用：比如淘宝的商品推荐，人浏览某个商品可以构建以下的异质图：用户输入查询，查询指向商品，同时用户点击商品。

参考如下论文：

• Beyond Smoothing: Unsupervised Graph Representation Learning with Edge Heterophily Discriminating, AAAI, [Paper]
• Restructuring Graph for Higher Homophily via Adaptive Spectral Clustering, AAAI, [Paper]
• 2-hop Neighbor Class Similarity (2NCS): A graph structural metric indicative of graph neural network performance, AAAI-W, [Paper]
• Ordered GNN: Ordering Message Passing to Deal with Heterophily and Over-smoothing, ICLR, [Paper]
• Gradient Gating for Deep Multi-Rate Learning on Graphs, ICLR, [Paper]
• ACMP: Allen-Cahn Message Passing with Attractive and Repulsive Forces for Graph Neural Networks, ICLR, [Paper]
• A Critical Look at Evaluation of GNNs Under Heterophily: Are We Really Making Progress?, ICLR, [Paper]
• GReTo: Remedying dynamic graph topology-task discordance via target homophily, ICLR, [Paper]
• Projections of Model Spaces for Latent Graph Inference, ICLR-W, [Paper]
• Addressing Heterophily in Graph Anomaly Detection: A Perspective of Graph Spectrum, WWW, [Paper]
• Homophily-oriented Heterogeneous Graph Rewiring, WWW, [Paper]
• Auto-HeG: Automated Graph Neural Network on Heterophilic Graphs, WWW, [Paper],
• Label Information Enhanced Fraud Detection against Low Homophily in Graphs, WWW, [Paper]
• SE-GSL: A General and Effective Graph Structure Learning Framework through Structural Entropy Optimization, WWW, [Paper]
• GCNH: A Simple Method For Representation Learning On Heterophilous Graphs, IJCNN, [Paper]
• Exploiting Neighbor Effect: Conv-Agnostic GNNs Framework for Graphs with Heterophily, TNNLS, [Paper]
• Multi-View Graph Representation Learning Beyond Homophily, TKDD, [Paper], [Code]
• Spatial Heterophily Aware Graph Neural Networks, KDD, [Paper]
• Finding the Missing-half: Graph Complementary Learning for Homophily-prone and Heterophily-prone Graphs, ICML, [Paper]
• Contrastive Learning Meets Homophily: Two Birds with One Stone, ICML, [Paper]
• Beyond Homophily: Reconstructing Structure for Graph-agnostic Clustering, ICML, [Paper]
• GOAT: A Global Transformer on Large-scale Graphs, ICML, [Paper]
• Towards Deep Attention in Graph Neural Networks: Problems and Remedies, ICML, [Paper]
• Half-Hop: A graph upsampling approach for slowing down message passing, ICML, [Paper]
• [Survey Paper] Heterophily and Graph Neural Networks: Past, Present and Future, Data Engineering, [Paper]
• Homophily-Enhanced Self-Supervision for Graph Structure Learning: Insights and Directions, TNNLS, [Paper]
• SlenderGNN: Accurate, Robust, and Interpretable GNN, and the Reasons for its Success, ICLR OpenReview, [Paper]
• Simple Spectral Graph Convolution from an Optimization Perspective, ICLR OpenReview, [Paper]
• The Impact of Neighborhood Distribution in Graph Convolutional Networks, ICLR OpenReview, [Paper]
• Wide Graph Neural Network, ICLR OpenReview, [Paper]
• Are Graph Attention Networks Attentive Enough? Rethinking Graph Attention by Capturing Homophily and Heterophily, ICLR OpenReview, [Paper]
• Node Classification Beyond Homophily: Towards a General Solution, ICLR OpenReview, [Paper]
• From ChebNet to ChebGibbsNet, ICLR OpenReview, [Paper]
• ProtoGNN: Prototype-Assisted Message Passing Framework for Non-Homophilous Graphs, ICLR OpenReview, [Paper]
• Low-Rank Graph Neural Networks Inspired by the Weak-balance Theory in Social Networks, ICLR OpenReview, [Paper]
• Can Single-Pass Contrastive Learning Work for Both Homophilic and Heterophilic Graph?, ICLR OpenReview, [Paper]
• Graph Contrastive Learning Under Heterophily: Utilizing Graph Filters to Generate Graph Views, ICLR OpenReview, [Paper]
• ReD-GCN: Revisit the Depth of Graph Convolutional Network, ICLR OpenReview, [Paper]
• Graph Neural Networks as Gradient Flows: Understanding Graph Convolutions via Energy, ICLR OpenReview, [Paper]
• Causally-guided Regularization of Graph Attention improves Generalizability, ICLR OpenReview, [Paper]
• GReTo: Remedying dynamic graph topology-task discordance via target homophily, ICLR OpenReview, [Paper]
• Statistical Mechanics of Generalization In Graph Convolution Networks, arXiv, [Paper]
• Semi-Supervised Classification with Graph Convolutional Kernel Machines, arXiv, [Paper]
• A Graph Neural Network with Negative Message Passing for Graph Coloring, arXiv, [Paper]
• Neighborhood Homophily-Guided Graph Convolutional Network, arXiv, [Paper]
• Is Signed Message Essential for Graph Neural Networks?, arXiv, [Paper]
• Attending to Graph Transformers, arXiv, [Paper]
• Heterophily-Aware Graph Attention Network, arXiv, [Paper]
• Steering Graph Neural Networks with Pinning Control, arXiv, [Paper]
• Contrastive Learning under Heterophily, arXiv, [Paper]
• Graph Positional Encoding via Random Feature Propagation, arXiv, [Paper]
• Truncated Affinity Maximization: One-class Homophily Modeling for Graph Anomaly Detection, arXiv, [Paper]
• When Do Graph Neural Networks Help with Node Classification: Investigating the Homophily Principle on Node Distinguishability, arXiv, [Paper]
• GPatcher: A Simple and Adaptive MLP Model for Alleviating Graph Heterophily, arXiv, [Paper]
• PathMLP: Smooth Path Towards High-order Homophily, arXiv, [Paper]
• Demystifying Structural Disparity in Graph Neural Networks: Can One Size Fit All?, arXiv, [Paper]
• Edge Directionality Improves Learning on Heterophilic Graphs, arXiv, [Paper]