At the beginning of 22, graph neural network (GNN) became a buzzword in the Internet circle. Throughout the year, the enthusiasm for research on GNN has been increasing, and it has become a research hotspot in major deep learning conferences.
GNN's excellent ability to process unstructured data has made new breakthroughs in network data analysis, recommendation systems, and natural language processing.
On the occasion of 23 years, this article reviews the latest research overview and research trends in the field of GNN in 22 years. I specially selected 5 representative articles to describe for you. I hope that students who want to publish papers in this field Bring some new ideas!
And give everyone a wave of benefits for free today !
Scan the QR code and reply to 【Graph Nerve】
to receive 5 papers in the article + 42 ICLR 2023 graph neural network papers
are all in pdf format, which is very convenient, and students who want it should not miss it!
“
01
Hyperbolic Graph Neural Network
topic:
This paper is a graph neural network based on a hyperbolic space, which builds a graph neural network in a hyperbolic space instead of our common Euclidean space. This research review is mainly by explaining what is a hyperbolic space? And why to establish a graph neural network in a hyperbolic space, and then study the advantages, applications, and current difficulties and opportunities of using a graph neural network in a hyperbolic space.
“
02
GNN-Based Graph Classification
research paper
topic:
A Survey of Graph Classification Research
Graph data widely exists in the real world, and can naturally represent complex associations between compound objects and their elements, but there is still a lack of a complete survey on graph classification research. This paper gives the definition of the graph classification problem and the challenges in this field; then combs and analyzes two types of graph classification methods—the graph classification method based on graph similarity calculation and the graph classification method based on graph neural network; and then gives the graph Evaluation indicators of classification methods, comparison of common data sets and experimental results; finally, common practical application scenarios of graph classification are introduced, future research directions in the field of graph classification are prospected, and the full text is summarized.
“
03
Federated Graph Machine Learning
topic:
As society increasingly focuses on data privacy, GNNs face the need to adapt to this new normal. This has led to the rapid development of federated graph neural network research in recent years. This paper proposes a new 3-layer taxonomy for federated graph neural networks to help researchers interested in this field understand how graph neural networks and federated learning complement each other. Finally, the article also looks forward to the future from 6 directions. Building more robust, dynamic, efficient and interpretable FedGNNs.
“
04
Equivariate graph neural network
topic:
This paper introduces the basic concepts in the equivariant graph neural network, and combines the literature published in the field of drug research and development to analyze and discuss the great application value of the equivariant graph neural network. The author analyzes the existing methods and divides them into three groups. to learn how to represent message passing and aggregation in GNNs. Benchmarks and related datasets are also summarized to facilitate later research method development and experimental evaluation. An outlook on potential future directions is also provided.
“
05
graph with heterogeneity
graph neural network
topic:
This paper provides a comprehensive overview of GNNs for heterogeneous graphs for the first time. Specifically, the authors propose a systematic taxonomy that essentially governs existing heterogeneous GNN models and provide a general summary and detailed analysis. And this paper summarizes mainstream heterogeneous graph benchmarks to facilitate robust and fair evaluation. Finally, the authors point out potential directions to advance and motivate future research and applications on heterogeneous graphs.
Send a wave of benefits to everyone for free today !
Scan the QR code and reply to 【Graph Nerve】
to receive 5 papers in the article + 42 ICLR 2023 graph neural network papers
are all in pdf format, which is very convenient, and students who want it should not miss it!
