Dec 9, 2022 · Inspired by doc2vec, a successful and efficient model in Natural Language Processing, graph embedding uses rooted subgraph and topological features to learn representations of graphs. Then, we can easily build a Machine Learning model to classify them.

Inspired by doc2vec, a successful and efficient model in Natural Lan-guage Processing, graph embedding uses rooted subgraph and topological fea-tures to learn representations of graphs. Then,...

Nov 7, 2024 · In this paper, we propose a novel method named \textit {Global-augmented Graph Structure Learning} (GaGSL), guided by the Graph Information Bottleneck (GIB) principle. The key idea behind GaGSL is to learn a compact and informative graph structure for node classification tasks.

Jun 3, 2021 · To address this issue, we present the CurGraph (Curriculum Learning for Graph Classification) framework, that analyzes the graph difficulty in the high-level semantic feature space. Specifically, we use the infomax method to obtain graph-level embeddings and a neural density estimator to model the embedding distributions.

Dec 15, 2020 · In order to avoid frequent subgraph mining and consider both the node-level and global features of the graph, a novel graph classification algorithm based on graph structure embedding is proposed from the perspective of graph feature vector construction.

Jun 13, 2024 · We propose a novel Motif-driven Subgraph Structure Learning method for Graph Classification (MOSGSL). Specifically, MOSGSL incorporates a subgraph structure learning module which can adaptively select important subgraphs.

Apr 1, 2023 · We propose a novel long-tailed GNN via graph structure learning (LTSL-GNN) that jointly learns graph structure and enhances graph embedding in an alternative way, which reduces the gap between head and tail nodes and significantly mitigates the …

To this end, we have conducted a comprehensive study on three key graph properties: homophily, degree distribution, and connected components, and determined how these factors influence semi-supervised node classification tasks. Specifically, the influence of homophily on GNN performance is rigorously assessed.

We have developed a comprehensive graph structure learning benchmark (GSLB), which consists of diverse graph datasets and state-of-the-art GSL algorithm. Feel free to explore our benchmark and provide any feedback or contributions. If you have come across relevant resources, feel free to open an issue or submit a pull request.

ich is graph structure learning for GNNs in sec-tion 14.3. This part will cover various topics including joint graph structure and representation learning for both unweighted and weighted graphs (section 14.3.1), and the connections to other problems such as graph generation, gr.