Recent advances in network embedding has revolutionized the field of graph and mining. However, (pre-)training embeddings for very large-scale networks is computationally challenging most existing methods. In this work, we present ProNE---a fast, scalable, effective model, whose single-thread version 10--400x faster than efficient benchmarks with 20 threads, including LINE, DeepWalk, node2vec, GraRep, HOPE. As a concrete example, single-version ProNE requires only 29 hours to embed hundreds...

We investigate how generative adversarial nets (GANs) can help semi-supervised learning on graphs. first provide insights working principles of over graphs and then present GraphSGAN, a novel approach to In generator classifier networks play competitive game. At equilibrium, generates fake samples in low-density areas between subgraphs. order discriminate from the real, implicitly takes density property subgraph into consideration. An efficient algorithm has been developed improve...

Linking entities from different sources is a fundamental task in building open knowledge graphs. Despite much research conducted related fields, the challenges of linkinglarge-scale heterogeneous entity graphs are far resolved. Employing two billion-scale academic (Microsoft Academic Graph and AMiner) as for our study, we propose unified framework --- LinKG to address problem large-scale linked graph. coupled with three linking modules, each which addresses one category entities. To link...

We study the problem of semi-supervised learning on graphs, for which graph neural networks (GNNs) have been extensively explored. However, most existing GNNs inherently suffer from limitations over-smoothing, non-robustness, and weak-generalization when labeled nodes are scarce. In this paper, we propose a simple yet effective framework -- GRAPH RANDOM NEURAL NETWORKS (GRAND) to address these issues. GRAND, first design random propagation strategy perform data augmentation. Then leverage...

Existing network embedding algorithms based on generative adversarial networks (GANs) improve the robustness of node embeddings by selecting high-quality negative samples with generator to play against discriminator. Since most can be easily discriminated from positive in graphs, their poor competitiveness weakens function generator. Inspired sales skills market, this article, we present tripartite training for (TriATNE), a novel learning framework stable and robust embeddings. TriATNE...

Link prediction in signed social networks is challenging because of the existence and imbalance three kinds status (positive, negative no-relation). Furthermore, there are a variety types no-relation reality, e.g., strangers frenemies, which cannot be well distinguished from other linked by existing approaches. In this paper, we propose novel Framework Integrating both Latent Explicit features (FILE), to better deal with improve overall link performance networks. particular, design two...

Graph data mining has largely benefited from the recent developments of graph representation learning. Most attempts to improve representations have thus far focused on designing new network embedding or neural (GNN) architectures. Inspired by SGC and ProNE models, we instead focus enhancing any existing learned further smoothing them via filters. In this paper, introduce an automated framework AutoProNE achieve this. Specifically, automatically searches for a unique optimal set filters...

Past research has examined how well these models grasp code syntax, yet their understanding of semantics still needs to be explored. We extensively analyze seven investigate represent syntax and semantics. This includes four prominent pre-trained (CodeBERT, GraphCodeBERT, CodeT5, UnixCoder) three large language (StarCoder, CodeLlama, CodeT5+). have developed probing tasks evaluate the models' abilities learn These focus on reconstructing semantic structures-such as AST, CFG, CDG, DDG -...

As the popularity of graph data increases, there is a growing need to count occurrences subgraph patterns interest, for variety applications. Many graphs are massive in scale and also fully dynamic (with insertions deletions edges), rendering exact computation these counts be infeasible. Common practice is, instead, use small set edges as sample estimate counts. Existing sampling algorithms with uniform probability. In this paper, we show that can do much better if based on their individual...

Training graph neural networks (GNNs) on large-scale data holds immense promise for numerous real-world applications but remains a great challenge. Several disk-based GNN systems have been built to train graphs in single machine. However, they often fall short terms of performance, especially when training terabyte-scale graphs. This is because existing either overly focus minimizing the number SSD accesses or do not fully overlap with training, thus resulting substantial unnecessary...

Graph neural networks are a promising architecture for learning and inference with graph-structured data. Yet, how to generate informative, fixed dimensional features graphs varying size topology can still be challenging. Typically, this is achieved through graph-pooling, which summarizes graph by compressing all its nodes into single vector. Is such “collapsing-style” graph-pooling the only choice classification? From complex system’s point of view, properties system arise largely from...

Network embedding aims to learn low-dimensional representations of nodes in a network, while the network structure and inherent properties are preserved. It has attracted tremendous attention recently due significant progress downstream learning tasks, such as node classification, link prediction, visualization. However, most existing methods suffer from expensive computations large volume networks. In this paper, we propose $10\times \sim 100\times$ faster method, called Progle, by...

Graph embedding has been widely used to process various downstream tasks on large-scale graphs, i.e. node classification, community detection and link prediction. Among methods, Deep Infomax (DGI) is a newly proposed method which achieves excellent performance in classification. However, such outstanding achievement may cause the over-mining issue of user privacy robustness this still unexplored. In paper, we investigate how disturb classification accuracy DGI from an attacker's perspective....

Recent advancements in Graph Neural Networks have led to state-of-the-art performance on graph representation learning. However, the majority of existing works process directed graphs by symmetrization, which causes loss directional information. To address this issue, we introduce magnetic Laplacian, a discrete Schr\"odinger operator with field, preserves edge directionality encoding it into complex phase an electric charge parameter. By adopting truncated variant PageRank named Linear-...

<div>Graph Convolutional Networks (GCNs) derive in spiration from recent advances computer vision, by stacking layers of first-order filters followed a nonlinear activation function to learn graph representations. Although GCNs have been shown boost the performance many network analysis tasks, they still face tremendous challenges learning Heterogeneous Information (HINs), where relations play decisive role knowledge reasoning. In addition, there may exist multi-aspect representations...

Graph Convolutional Networks (GCNs) derive inspiration from recent advances in computer vision, by stacking layers of first-order filters followed a nonlinear activation function to learn entity or graph embeddings. Although GCNs have been shown boost the performance many network analysis tasks, they still face tremendous challenges learning Heterogeneous Information (HINs), where relations play decisive role knowledge reasoning. What's more, there are multiaspect representations entities...