Graph Machine Learning is essential for understanding and analyzing relational data. However, privacy-sensitive applications demand the ability to efficiently remove sensitive information from trained graph neural networks (GNNs), avoiding unnecessary time space overhead caused by retraining models scratch. To address this issue, Unlearning (GU) has emerged as a critical solution, with potential support dynamic updates in data management systems enable scalable unlearning distributed while...

The $q$-parameterized magnetic Laplacian serves as the foundation of directed graph (digraph) convolution, enabling this kind digraph neural network (MagDG) to encode node features and structural insights by complex-domain message passing. As a generalization undirected methods, MagDG shows superior capability in modeling intricate web-scale topology. Despite great success achieved existing MagDGs, limitations still exist: (1) Hand-crafted $q$: performance MagDGs depends on selecting an...

Machine unlearning, as a pivotal technology for enhancing model robustness and data privacy, has garnered significant attention in prevalent web mining applications, especially thriving graph-based scenarios. However, most existing graph unlearning (GU) approaches face challenges due to the intricate interactions among web-scale elements during training: (1) The gradient-driven node entanglement hinders complete knowledge removal response requests; (2) billion-level scenarios present...

Federated graph learning (FGL) has emerged as a promising paradigm for collaborative machine learning, enabling multiple parties to jointly train models while preserving the privacy of raw data. However, existing FGL methods often overlook model-centric heterogeneous (MHtFGL) problem, which arises in real-world applications, such aggregation from different companies with varying scales and architectures. MHtFGL presents an additional challenge: diversity client model architectures hampers...

Dynamic graphs (DGs), which capture time-evolving relationships between graph entities, have widespread real-world applications. To efficiently encode DGs for downstream tasks, most dynamic neural networks follow the traditional message-passing mechanism and extend it with time-based techniques. Despite their effectiveness, growth of historical interactions introduces significant scalability issues, particularly in industry scenarios. address this limitation, we propose ScaDyG, core idea...

The embedded representation and clustering tasks both play important roles in relational data analysis mining. Traditional methods mainly employ graph structure to describe data, but intuitive pairwise connections among nodes are insufficient model high-order the real-world, such as relations between proteins polypeptide chains. Hypergraphs a generalization of graphs, hypergraphs can well data. When modeling real world, often accompanied by node attributes, i.e. attributed hypergraphs....

Graph clustering is a fundamental task in data analysis and has attracted considerable attention recommendation systems, mapping knowledge domain, biological science. Because graph convolution very effective combining the feature information topology of data, some methods based on have achieved superior performance. However, current lack consideration structured process convolution. Specifically, most existing ignore implicit interaction between information, stacking small number...

Federated Graph Learning (FGL) is a distributed machine learning paradigm that enables collaborative training on large-scale subgraphs across multiple local systems. Existing FGL studies fall into two categories: (i) Optimization, which improves multi-client in existing models; (ii) Model, enhances performance with complex models and interactions. However, most optimization strategies are designed specifically for the computer vision domain ignore graph structure, presenting dissatisfied...

Recently, recommender systems based on Graph Convolution Network (GCN) have become a research hotspot, especially in collaborative filtering. However, most GCN-based models inferior embedding propagation mechanism, leading to low information extraction efficiency. Besides, the existing methods suffer from high computational complexity for large user-item interaction graphs. In order solve above problems, we propose LII-GCCF that integrates Linear transformation, Initial residual and Identity...

Most existing graph neural networks (GNNs) are limited to undirected graphs, whose restricted scope of the captured relational information hinders their expressive capabilities and deployments in real-world scenarios. Compared with directed graphs (digraphs) fit demand for modeling more complex topological systems by capturing intricate relationships between nodes, such as formulating transportation financial networks. While some GNNs have been introduced, inspiration mainly comes from deep...

With the rapid advancement of AI applications, growing needs for data privacy and model robustness have highlighted importance machine unlearning, especially in thriving graph-based scenarios. However, most existing graph unlearning strategies primarily rely on well-designed architectures or manual process, rendering them less user-friendly posing challenges terms deployment efficiency. Furthermore, striking a balance between performance framework generalization is also pivotal concern. To...

Federated Graph Learning (FGL) is a distributed machine learning paradigm that enables collaborative training on large-scale subgraphs across multiple local systems. Existing FGL studies fall into two categories: (i) Optimization, which improves multi-client in existing models; (ii) Model, enhances performance with complex models and interactions. However, most optimization strategies are designed specifically for the computer vision domain ignore graph structure, presenting dissatisfied...

Recently, Federated Graph Learning (FGL) has attracted significant attention as a distributed framework based on graph neural networks, primarily due to its capability break data silos. Existing FGL studies employ community split the homophilous global by default simulate federated semi-supervised node classification settings. Such strategy assumes consistency of topology between multi-client subgraphs and graph, where connected nodes are highly likely possess similar feature distributions...

Scalable graph neural networks (GNNs) have emerged as a promising technique, which exhibits superior predictive performance and high running efficiency across numerous large-scale graph-based web applications. However, (i) Most scalable GNNs tend to treat all nodes in graphs with the same propagation rules, neglecting their topological uniqueness; (ii) Existing node-wise optimization strategies are insufficient on web-scale intricate topology, where full portrayal of nodes' local properties...

With the rapid advancement of AI applications, growing needs for data privacy and model robustness have highlighted importance machine unlearning, especially in thriving graph-based scenarios. However, most existing graph unlearning strategies primarily rely on well-designed architectures or manual process, rendering them less user-friendly posing challenges terms deployment efficiency. Furthermore, striking a balance between performance framework generalization is also pivotal concern. To...

Subgraph federated learning (subgraph-FL) is a new distributed paradigm that facilitates the collaborative training of graph neural networks (GNNs) by multi-client subgraphs. Unfortunately, significant challenge subgraph-FL arises from subgraph heterogeneity, which stems node and topology variation, causing impaired performance global GNN. Despite various studies, they have not yet thoroughly investigated impact mechanism heterogeneity. To this end, we decouple revealing correspond to...

Graph Neural Networks (GNNs) have demonstrated effectiveness in various graph-based tasks. However, their inefficiency training and inference presents challenges for scaling up to real-world large-scale graph applications. To address the critical challenges, a range of algorithms been proposed accelerate GNNs, attracting increasing attention from research community. In this paper, we present systematic review acceleration which can be categorized into three main topics based on purpose:...

Most existing graph neural networks (GNNs) are limited to undirected graphs, whose restricted scope of the captured relational information hinders their expressive capabilities and deployment. Compared with directed graphs (digraphs) fit demand for modeling more complex topological systems by capturing intricate relationships between nodes. While some GNNs have been introduced, inspiration mainly comes from deep learning architectures, which lead redundant complexity computation, making them...