49篇ICLR2020高分「图机器学习GML」接受论文及代码

2020 年 1 月 18 日 专知

【导读】ICLR2020论文收到2594篇论文提交，有687篇被接受，接受率为26.5%。在关于图机器学习方面，Sergei Ivanov整理了关于图机器学习方面的高分论文，有49篇关于图机器学习论文，专知进一步整理了论文阅读欢迎查看！

总共有152篇论文在题目中包含了一个“Graph”，其中有49篇论文被接受。所有图表论文的平均评分是4.5，而被接受的论文的平均评分是6.3。

1 图神经网络的逻辑表达性 The Logical Expressiveness of Graph Neural Networks

Pablo Barceló, Egor V. Kostylev, Mikael Monet, Jorge Pérez, Juan Reutter, Juan Pablo Silva

代码地址：

https://anonymous.4open.science/r/787222e2-ad5e-4810-a788-e80f0fe7eff0/

论文地址：

https://openreview.net/forum?id=r1lZ7AEKvB

本文重点研究了图神经网络的表达特性。审稿人对作者充分且一致地回答了他们的问题表示满意，认为这是一篇应该被接受的强有力的论文

2 Hyper-SAGNN:一种基于自注意力的超图神经网络 Hyper-SAGNN: a self-attention based graph neural network for hypergraphs

Ruochi Zhang, Yuesong Zou, Jian Ma

代码地址：

https://drive.google.com/drive/folders/1kIOc4SlAJllUJsrr2OnZ4izIQIw2JexU?usp=sharing

论文地址：

https://openreview.net/forum?id=ryeHuJBtPH

本文介绍了一种新的神经网络模型，该模型可以表示可变尺寸的超边缘，并通过实验证明了该模型在一些问题上可以改进或匹配目前的技术水平。

3. 论节点嵌入与结构图表示的等价性，On the Equivalence between Node Embeddings and Structural Graph Representations

Balasubramaniam Srinivasan, Bruno Ribeiro

论文地址：

https://openreview.net/forum?id=SJxzFySKwH

本文给出了节点嵌入与结构图表示之间的关系。通过对结构节点表示的含义和节点嵌入的含义的仔细定义，利用置换群，作者在定理2中证明了节点嵌入不能表示结构表示中没有的任何额外信息。然后，本文对三个任务进行了实证实验，并在第四个任务中对理论结果进行了说明。

4. LambdaNet:使用图神经网络的概率类型推断，LambdaNet: Probabilistic Type Inference using Graph Neural Networks

Jiayi Wei, Maruth Goyal, Greg Durrett, Isil Dillig

论文地址：

https://openreview.net/forum?id=Hkx6hANtwH

本文提出了一种基于图神经网络的动态语言类型推理方法。Reviewer(以及区域主席)喜欢GNNs在实际问题、演示和结果中的这种新颖而有用的应用。明确的接受。

5. 定向消息传递分子图，Directional Message Passing for Molecular Graphs

Johannes Klicpera, Janek Groß, Stephan Günnemann

论文地址：

https://openreview.net/forum?id=B1eWbxStPH

本文研究了量子化学的图神经网络，在此基础上加入了一些物理方面的创新。特别地，它在保持等方差的同时考虑方向边缘信息。

49篇完整「图机器学习GML」论文列表：

论文题目	论文地址	接受类型	打分
The Logical Expressiveness of Graph Neural Net...	https://openreview.net/forum?id=r1lZ7AEKvB	Accept (Spotlight)	8
Hyper-SAGNN: a self-attention based graph neur...	https://openreview.net/forum?id=ryeHuJBtPH	Accept (Poster)	8
On the Equivalence between Node Embeddings and...	https://openreview.net/forum?id=SJxzFySKwH	Accept (Poster)	7.33
LambdaNet: Probabilistic Type Inference using ...	https://openreview.net/forum?id=Hkx6hANtwH	Accept (Poster)	7.33
Directional Message Passing for Molecular Graphs	https://openreview.net/forum?id=B1eWbxStPH	Accept (Spotlight)	7.33
What graph neural networks cannot learn: depth...	https://openreview.net/forum?id=B1l2bp4YwS	Accept (Poster)	7.33
Graph Neural Networks Exponentially Lose Expre...	https://openreview.net/forum?id=S1ldO2EFPr	Accept (Spotlight)	7.33
Physics-aware Difference Graph Networks for Sp...	https://openreview.net/forum?id=r1gelyrtwH	Accept (Poster)	7.33
GLAD: Learning Sparse Graph Recovery	https://openreview.net/forum?id=BkxpMTEtPB	Accept (Poster)	7.33
GraphZoom: A Multi-level Spectral Approach for...	https://openreview.net/forum?id=r1lGO0EKDH	Accept (Talk)	7.33
DeepSphere: a graph-based spherical CNN	https://openreview.net/forum?id=B1e3OlStPB	Accept (Spotlight)	6.67
Query2box: Reasoning over Knowledge Graphs in ...	https://openreview.net/forum?id=BJgr4kSFDS	Accept (Poster)	6.67
Dynamically Pruned Message Passing Networks fo...	https://openreview.net/forum?id=rkeuAhVKvB	Accept (Poster)	6.67
Inductive Matrix Completion Based on Graph Neu...	https://openreview.net/forum?id=ByxxgCEYDS	Accept (Spotlight)	6.67
On the geometry and learning low-dimensional e...	https://openreview.net/forum?id=SkxQp1StDH	Accept (Poster)	6.67
Reinforced Genetic Algorithm Learning for Opti...	https://openreview.net/forum?id=rkxDoJBYPB	Accept (Poster)	6.67
Reinforcement Learning Based Graph-to-Sequence...	https://openreview.net/forum?id=HygnDhEtvr	Accept (Poster)	6.67
Learning to Retrieve Reasoning Paths over Wiki...	https://openreview.net/forum?id=SJgVHkrYDH	Accept (Poster)	6.67
A Fair Comparison of Graph Neural Networks for...	https://openreview.net/forum?id=HygDF6NFPB	Accept (Poster)	6.67
You CAN Teach an Old Dog New Tricks! On Traini...	https://openreview.net/forum?id=BkxSmlBFvr	Accept (Poster)	6.67
Geom-GCN: Geometric Graph Convolutional Networks	https://openreview.net/forum?id=S1e2agrFvS	Accept (Spotlight)	6.67
HOPPITY: LEARNING GRAPH TRANSFORMATIONS TO DET...	https://openreview.net/forum?id=SJeqs6EFvB	Accept (Spotlight)	6.67
Inductive representation learning on temporal ...	https://openreview.net/forum?id=rJeW1yHYwH	Accept (Poster)	6.67
Measuring and Improving the Use of Graph Infor...	https://openreview.net/forum?id=rkeIIkHKvS	Accept (Poster)	6.33
Memory-Based Graph Networks	https://openreview.net/forum?id=r1laNeBYPB	Accept (Poster)	6
Adaptive Structural Fingerprints for Graph Att...	https://openreview.net/forum?id=BJxWx0NYPr	Accept (Poster)	6
Graph Constrained Reinforcement Learning for N...	https://openreview.net/forum?id=B1x6w0EtwH	Accept (Poster)	6
Graph Convolutional Reinforcement Learning	https://openreview.net/forum?id=HkxdQkSYDB	Accept (Poster)	6
Inductive and Unsupervised Representation Lear...	https://openreview.net/forum?id=rkem91rtDB	Accept (Poster)	6
StructPool: Structured Graph Pooling via Condi...	https://openreview.net/forum?id=BJxg_hVtwH	Accept (Poster)	6
GraphSAINT: Graph Sampling Based Inductive Lea...	https://openreview.net/forum?id=BJe8pkHFwS	Accept (Poster)	6
Physics-as-Inverse-Graphics: Unsupervised Phys...	https://openreview.net/forum?id=BJeKwTNFvB	Accept (Poster)	6
Strategies for Pre-training Graph Neural Networks	https://openreview.net/forum?id=HJlWWJSFDH	Accept (Spotlight)	6
Deep Graph Matching Consensus	https://openreview.net/forum?id=HyeJf1HKvS	Accept (Poster)	6
InfoGraph: Unsupervised and Semi-supervised Gr...	https://openreview.net/forum?id=r1lfF2NYvH	Accept (Spotlight)	6
Pruned Graph Scattering Transforms	https://openreview.net/forum?id=rJeg7TEYwB	Accept (Poster)	6
Curvature Graph Network	https://openreview.net/forum?id=BylEqnVFDB	Accept (Poster)	6
Composition-based Multi-Relational Graph Convo...	https://openreview.net/forum?id=BylA_C4tPr	Accept (Poster)	6
GraphAF: a Flow-based Autoregressive Model for...	https://openreview.net/forum?id=S1esMkHYPr	Accept (Poster)	6
Graph inference learning for semi-supervised c...	https://openreview.net/forum?id=r1evOhEKvH	Accept (Poster)	6
Probability Calibration for Knowledge Graph Em...	https://openreview.net/forum?id=S1g8K1BFwS	Accept (Poster)	5.75
Neural Execution of Graph Algorithms	https://openreview.net/forum?id=SkgKO0EtvS	Accept (Poster)	5.67
Visual Representation Learning with 3D View-Co...	https://openreview.net/forum?id=BJxt60VtPr	Accept (Poster)	5.25
Abstract Diagrammatic Reasoning with Multiplex...	https://openreview.net/forum?id=ByxQB1BKwH	Accept (Poster)	5
FEW-SHOT LEARNING ON GRAPHS VIA SUPER-CLASSES ...	https://openreview.net/forum?id=Bkeeca4Kvr	Accept (Poster)	5
Differentiable learning of numerical rules in ...	https://openreview.net/forum?id=rJleKgrKwS	Accept (Poster)	5
Learning deep graph matching with channel-inde...	https://openreview.net/forum?id=rJgBd2NYPH	Accept (Poster)	5
DropEdge: Towards Deep Graph Convolutional Net...	https://openreview.net/forum?id=Hkx1qkrKPr	Accept (Poster)	4
Efficient Probabilistic Logic Reasoning with G...	https://openreview.net/forum?id=rJg76kStwH	Accept (Poster)	2.33