Topology-imbalance is a graph-specific imbalance problem caused by the uneven topology positions of labeled nodes, which significantly damages performance GNNs. What topology-imbalance means and how to measure its impact on graph learning remain under-explored. In this paper, we provide new understanding from global view supervision information distribution in terms under-reaching over-squashing, motivates two quantitative metrics as measurements. light our analysis, propose novel...

Learning unbiased node representations for imbalanced samples in the graph has become a more remarkable and important topic. For graph, significant challenge is that topological properties of nodes (e.g., locations, roles) are unbalanced (topology-imbalance), other than number training labeled (quantity-imbalance). Existing studies on topology-imbalance focus location or local neighborhood structure nodes, ignoring global underlying hierarchical i.e., hierarchy. In real-world scenario, data...

Wearable Human Activity Recognition (WHAR) is a prominent research area within ubiquitous computing. Multi-sensor synchronous measurement has proven to be more effective for WHAR than using single sensor. However, existing methods use shared convolutional kernels indiscriminate temporal feature extraction across each sensor variable, which fails effectively capture spatio-temporal relationships of intra-sensor and inter-sensor variables. We propose the DecomposeWHAR model consisting...

Adversarial evasion attacks pose significant threats to graph learning, with lines of studies that have improved the robustness Graph Neural Networks (GNNs). However, existing works rely on priors about clean graphs or attacking strategies, which are often heuristic and inconsistent. To achieve robust learning over different types diverse datasets, we investigate this problem from a prior-free structure purification perspective. Specifically, propose novel Diffusion-based Structure...

Graph neural networks(GNNs) have been demonstrated to depend on whether the node effective information is sufficiently passing. Discrete curvature (Ricci curvature) used study graph connectivity and propagation efficiency with a geometric perspective, has raised in recent years explore efficient message-passing structure of GNNs. However, most empirical studies are based directly observed structures or heuristic topological assumptions, lack in-depth exploration underlying optimal transport...

Wearable Human Activity Recognition (WHAR) is a prominent research area within ubiquitous computing. Multi-sensor synchronous measurement has proven to be more effective for WHAR than using single sensor. However, existing methods use shared convolutional kernels indiscriminate temporal feature extraction across each sensor variable, which fails effectively capture spatio-temporal relationships of intra-sensor and inter-sensor variables. We propose the DecomposeWHAR model consisting...

Real-world graphs have inherently complex and diverse topological patterns, known as heterogeneity. Most existing works learn graph representation in a single constant curvature space that is insufficient to match the geometric shapes, resulting low-quality embeddings with high distortion. This also constitutes critical challenge for foundation models, which are expected uniformly handle wide variety of data. Recent studies indicated product manifold gains possibility address However, still...

Dynamic graphs exhibit intertwined spatio-temporal evolutionary patterns, widely existing in the real world. Nevertheless, structure incompleteness, noise, and redundancy result poor robustness for Graph Neural Networks (DGNNs). Structure Learning (DGSL) offers a promising way to optimize graph structures. However, aside from encountering unacceptable quadratic complexity, it overly relies on heuristic priors, making hard discover underlying predictive patterns. How efficiently refine...

Hierarchy is an important and commonly observed topological property in real-world graphs that indicate the relationships between supervisors subordinates or organizational behavior of human groups. As hierarchy introduced as a new inductive bias into Graph Neural Networks (GNNs) various tasks, it implies latent relations for attackers to improve their inference attack performance, leading serious privacy leakage issues. In addition, existing privacy-preserving frameworks suffer from reduced...

Dynamic graph neural networks (DGNNs) are increasingly pervasive in exploiting spatio-temporal patterns on dynamic graphs. However, existing works fail to generalize under distribution shifts, which common real-world scenarios. As the generation of graphs is heavily influenced by latent environments, investigating their impacts out-of-distribution (OOD) generalization critical. it remains unexplored with following two major challenges: (1) How properly model and infer complex environments...

Dynamic Graphs widely exist in the real world, which carry complicated spatial and temporal feature patterns, challenging their representation learning. Graph Neural Networks (DGNNs) have shown impressive predictive abilities by exploiting intrinsic dynamics. However, DGNNs exhibit limited robustness, prone to adversarial attacks. This paper presents novel Information Bottleneck (DGIB) framework learn robust discriminative representations. Leveraged (IB) principle, we first propose expected...

Summary This article addresses the issue of ensuring model accuracy and training efficiency in a constrained federated learning environment. In an actual environment, each device's software, hardware, network conditions are heterogeneous. Some terminal devices may not be able to undertake work assigned by server, resulting poor slower convergence speed. However, existing research cannot ensure that device participating can handle workload allocated system without collecting too much...

Deep graph learning has gained grand popularity over the past years due to its versatility and success in representing data across a wide range of domains. However, pervasive issue imbalanced distributions, where certain parts exhibit disproportionally abundant while others remain sparse, undermines efficacy conventional algorithms, leading biased outcomes. To address this challenge, Imbalanced Graph Learning (IGL) garnered substantial attention, enabling more balanced distributions better...

Graph neural networks(GNNs) have been demonstrated to depend on whether the node effective information is sufficiently passing. Discrete curvature (Ricci curvature) used study graph connectivity and propagation efficiency with a geometric perspective, has raised in recent years explore efficient message-passing structure of GNNs. However, most empirical studies are based directly observed structures or heuristic topological assumptions lack in-depth exploration underlying optimal transport...

Dynamic graphs exhibit intertwined spatio-temporal evolutionary patterns, widely existing in the real world. Nevertheless, structure incompleteness, noise, and redundancy result poor robustness for Graph Neural Networks (DGNNs). Structure Learning (DGSL) offers a promising way to optimize graph structures. However, aside from encountering unacceptable quadratic complexity, it overly relies on heuristic priors, making hard discover underlying predictive patterns. How efficiently refine...

Real-world graphs have inherently complex and diverse topological patterns, known as heterogeneity. Most existing works learn graph representation in a single constant curvature space that is insufficient to match the geometric shapes, resulting low-quality embeddings with high distortion. This also constitutes critical challenge for foundation models, which are expected uniformly handle wide variety of data. Recent studies indicated product manifold gains possibility address However, still...

Hierarchy is an important and commonly observed topological property in real-world graphs that indicate the relationships between supervisors subordinates or organizational behavior of human groups. As hierarchy introduced as a new inductive bias into Graph Neural Networks (GNNs) various tasks, it implies latent relations for attackers to improve their inference attack performance, leading serious privacy leakage issues. In addition, existing privacy-preserving frameworks suffer from reduced...