ICLR'23上的高分GNN论文. 来看看你的投稿得分是啥水平吧

今年ICLR所有投稿论文均分为4.94，论文接收分数大概在5.5分，该文章主要收集录取概率大（6分以上）的GNN相关论文，6分以上的论文排名为前20%（共~1000篇），其中GNN相关论文占比约4%（共~43篇）。

7分以上的GNN论文

6分以上的GNN论文汇总：

论文标题后面的表示分数情况，格式为[mean±std: scores]

Contrastive Learning

Localized Graph Contrastive Learning [6.00±1.22: 5;8;6;5]

Graph Contrastive Learning for Skeleton-based Action Recognition [6.00±2.12: 5;8;3;8]

Pre-Training

Mole-BERT: Rethinking Pre-training Graph Neural Networks for Molecules [6.25±1.09: 6;8;6;5]

Expressivity

Boosting the Cycle Counting Power of Graph Neural Networks with I -GNNs [6.60±1.20: 8;5;6;6;8]

Cycle to Clique (Cy2C) Graph Neural Network: A Sight to See beyond Neighborhood Aggregation [6.33±1.25: 8;6;5]

Graph Attention vs. Convolution Learnable Graph Convolutional Attention Networks [6.33±1.25: 5;6;8]

Graph Classification

Automated Data Augmentations for Graph Classification [7.00±1.41: 5;8;8]

Link Prediction

Graph Neural Networks for Link Prediction with Subgraph Sketching [8.50±0.87：8;8;8;10]

MetaGL: Evaluation-Free Selection of Graph Learning Models via Meta-Learning [6.25±1.09: 6;6;5;8]

Imbalanced Classification

Diving into Unified Data-Model Sparsity for Class-Imbalanced Graph Representation Learning [6.33±2.36: 3;8;8]

Robustness

Learning MLPs on Graphs: A Unified View of Effectiveness, Robustness, and Efficiency [6.75±1.30: 6;8;8;5]

Chasing All-Round Graph Representation Robustness: Model, Training, and Optimization [6.75±2.17: 8;3;8;8]

Scalability

LMC: Fast Training of GNNs via Subgraph Sampling with Provable Convergence [6.25±2.05: 8;8;6;3]

Distributed Graph Neural Network Training with Periodic Stale Representation Synchronization [6.00±1.22: 6;5;8;5]

AutoML

AutoGT: Automated Graph Transformer Architecture Search [7.33±0.94:8;8;6]

AutoTransfer: AutoML with Knowledge Transfer - An Application to Graph Neural Networks [6.67±0.94: 8;6;6]

Fairness

Learning Fair Graph Representations via Automated Data Augmentations [7.00±1.00: 8;8;6;6]

Equivariance

Conditional Antibody Design as 3D Equivariant Graph Translation [8.00±0.00: 8;8;8;8]

Anisotropic Message Passing: Graph Neural Networks with Directional and Long-Range Interactions [6.25±1.09: 6;6;8;5]

Equiformer: Equivariant Graph Attention Transformer for 3D Atomistic Graphs [6.00±1.22: 5;5;6;8]

Generative Models

DiGress: Discrete Denoising diffusion for graph generation [6.67±0.94: 8;6;6]

Adversarial Attacks

Characterizing the Influence of Graph Elements [6.50±0.87: 6;6;8;6]

Generalisability

Control Graph as Unified IO for Morphology-Task Generalization [6.50±1.50: 5;8;8;5]

Multi-task Self-supervised Graph Neural Networks Enable Stronger Task Generalization [6.00±0.00: 6;6;6]

Energy-based Out-of-Distribution Detection for Graph Neural Networks [6.00±1.22: 5;5;8;6]

NLP

Learning on Large-scale Text-attributed Graphs via Variational Inference [7.25±1.30: 5;8;8;8]

Knowledge Graph

ExpressivE: A Spatio-Functional Embedding For Knowledge Graph Completion [7.25±1.92: 8;5;10;6]

Logical Entity Representation in Knowledge-Graphs for Differentiable Rule Learning [6.00±1.22: 8;5;6;5]

Multimodal Analogical Reasoning over Knowledge Graphs [6.00±1.41: 5;5;8]

Neural Compositional Rule Learning for Knowledge Graph Reasoning [6.00±2.12: 3;8;5;8]

Physics

Learning rigid dynamics with face interaction graph networks [7.00±1.73: 6;10;6;6]

Learning Symbolic Models for Graph-structured Physical Mechanism [6.00±1.41: 5;5;8]

Unlearning

Efficient Model Updates for Approximate Unlearning of Graph-Structured Data [6.67±0.94: 6;6;8]

Quadratic Programs

On Representing Linear Programs by Graph Neural Networks [6.33±1.25: 8;6;5]

Proteins

Matching receptor to odorant with protein language and graph neural networks [6.33±1.25: 6;8;5]

Energy Models

Learnable Topological Features For Phylogenetic Inference via Graph Neural Networks [6.33±2.36: 3;8;8]

Circuit

CktGNN: Circuit Graph Neural Network for Electronic Design Automation [6.25±1.09: 5;8;6;6]

Hypergraph

Ollivier-Ricci Curvature for Hypergraphs: A Unified Framework [6.25±1.09: 6;8;5;6]

Explainability

A Differential Geometric View and Explainability of GNN on Evolving Graphs [6.25±1.09: 8;6;6;5]

Temporal Data

GReTo: Remedying dynamic graph topology-task discordance via target homophily [6.00±1.10: 6;6;8;5;5]

Regularisation

Distributional Signals for Node Classification in Graph Neural Networks [6.00±1.41: 5;8;5]

Miscellaneous

Effects of Graph Convolutions in Multi-layer Networks [7.50±0.87: 8;8;8;6]

A Convergent Single-Loop Algorithm for Gromov-Wasserstein in Graph Data [7.25±1.30: 8;8;8;5]

Refs：

• https://github.com/fedebotu/ICLR2023-OpenReviewData
• github.com/fedebotu/ICLR2023-OpenReviewData Google文档-论文ratings
• docs.google.com/spreadsheets/d/1INZI9epkfBkPOlKuJFaffUCOKDns87Iqg4zovHnf-zs/edit
• https://guoqiangwei.xyz/iclr2023_stats/iclr2023_submissions.html
• guoqiangwei.xyz/iclr2023_stats/iclr2023_submissions.html
• ICLR 2023 Conference openreview.net/group?id=ICLR.cc/2023/Conference