Advancements in Graph Neural Networks

The field of graph neural networks (GNNs) is rapidly advancing, with recent developments focusing on improving the robustness, expressiveness, and efficiency of these models. One of the key directions is the incorporation of hierarchical and hypergraph structures, which enables better modeling of complex relationships and multi-scale interactions. Additionally, there is a growing interest in developing GNNs that can handle temporal and spatial dependencies, as well as those that can provide interpretable and explainable results. Another important area of research is the development of robust and reliable GNNs that can withstand adversarial attacks and Perturbations. Noteworthy papers in this area include the proposal of ChemHGNN, a hierarchical hypergraph neural network for reaction virtual screening and discovery, and KCES, a training-free defense framework for robust GNNs via kernel complexity. Overall, the field of GNNs is moving towards more advanced and specialized models that can tackle complex real-world problems.

Sources

Mini-Game Lifetime Value Prediction in WeChat

ChemHGNN: A Hierarchical Hypergraph Neural Network for Reaction Virtual Screening and Discovery

KCES: Training-Free Defense for Robust Graph Neural Networks via Kernel Complexity

Geometry-Aware Edge Pooling for Graph Neural Networks

TrustGLM: Evaluating the Robustness of GraphLLMs Against Prompt, Text, and Structure Attacks

Analysis of Anonymous User Interaction Relationships and Prediction of Advertising Feedback Based on Graph Neural Network

Logical Expressiveness of Graph Neural Networks with Hierarchical Node Individualization

Density-aware Walks for Coordinated Campaign Detection

Toward a Graph Foundation Model: Pre-Training Transformers With Random Walks

Evaluating Loss Functions for Graph Neural Networks: Towards Pretraining and Generalization

CLGNN: A Contrastive Learning-based GNN Model for Betweenness Centrality Prediction on Temporal Graphs

Equivariance Everywhere All At Once: A Recipe for Graph Foundation Models

sHGCN: Simplified hyperbolic graph convolutional neural networks

Optimization of bi-directional gated loop cell based on multi-head attention mechanism for SSD health state classification model

Over-squashing in Spatiotemporal Graph Neural Networks

Built with on top of