Graph Neural Networks (GNNs) have shown superior performance in analyzing attributed networks various web-based applications such as social recommendation and web search. Nevertheless, high-stake decision-making scenarios online fraud detection, there is an increasing societal concern that GNNs could make discriminatory decisions towards certain demographic groups. Despite recent explorations on fair GNNs, these works are tailored for a specific GNN model. However, myriads of variants been...

Graph mining algorithms have been playing a significant role in myriad fields over the years. However, despite their promising performance on various graph analytical tasks, most of these lack fairness considerations. As consequence, they could lead to discrimination towards certain populations when exploited human-centered applications. Recently, algorithmic has extensively studied graph-based In contrast independent and identically distributed (i.i.d.) data, exclusive backgrounds,...

Recent years have witnessed the pivotal role of Graph Neural Networks (GNNs) in various high-stake decision-making scenarios due to their superior learning capability. Close on heels successful adoption GNNs different application domains has been increasing societal concern that conventional often do not fairness considerations. Although some research progress made improve GNNs, these works mainly focus notion group regarding subgroups defined by a protected attribute such as gender, age,...

Graph Neural Networks (GNNs) have shown great power in learning node representations on graphs. However, they may inherit historical prejudices from training data, leading to discriminatory bias predictions. Although some work has developed fair GNNs, most of them directly borrow representation techniques non-graph domains without considering the potential problem sensitive attribute leakage caused by feature propagation GNNs. we empirically observe that could vary correlation previously...

Graph machine learning has gained great attention in both academia and industry recently. Most of the graph models, such as Neural Networks (GNNs), are trained over massive data. However, many realworld scenarios, hospitalization prediction healthcare systems, data is usually stored at multiple owners cannot be directly accessed by any other parties due to privacy concerns regulation restrictions. Federated Machine Learning (FGML) a promising solution tackle this challenge training models...

Graph Neural Networks (GNNs) have emerged as the leading paradigm for solving graph analytical problems in various real-world applications. Nevertheless, GNNs could potentially render biased predictions towards certain demographic subgroups. Understanding how bias arises is critical, it guides design of GNN debiasing mechanisms. However, most existing works overwhelmingly focus on debiasing, but fall short explaining such induced. In this paper, we study a novel problem interpreting...

Analyzing the causal impact of different policies in reducing spread COVID-19 is critical importance. The main challenge here existence unobserved confounders (e.g., vigilance residents) which influence both presence and COVID-19. Besides, as may be time-varying, it even more difficult to capture them. Fortunately, increasing prevalence web data from various online applications provides an important resource time-varying observational data, enhances opportunity them, e.g., residents over...

Graph Neural Networks have recently become a prevailing paradigm for various high-impact graph analytical problems. Existing efforts can be mainly categorized as spectral-based and spatial-based methods. The major challenge the former is to find an appropriate filter distill discriminative information from input signals learning. Recently, myriads of explorations are made achieve better filters, e.g., Convolutional Network (GCN), which leverages Chebyshev polynomial truncation seek...

Few-shot node classification aims at classifying nodes with limited labeled as references. Recent few-shot methods typically learn from classes abundant (i.e., meta-training classes) and then generalize to meta-test classes). Nevertheless, on real-world graphs, it is usually difficult obtain for many classes. In practice, each class can only consist of several nodes, known the extremely weak supervision problem. classification, meta-training, generalization gap between will become larger...

Self-supervised learning with masked autoencoders has recently gained popularity for its ability to produce effective image or textual representations, which can be applied various downstream tasks without retraining. However, we observe that the current autoencoder models lack good generalization on graph data. To tackle this issue, propose a novel framework called GiGaMAE. Different from existing learn node presentations by explicitly reconstructing original components (e.g., features...

With the wide adoption of mobile devices and web applications, location-based social networks (LBSNs) offer large-scale individual-level location-related activities experiences. Next point-of-interest (POI) recommendation is one most important tasks in LBSNs, aiming to make personalized recommendations next suitable locations users by discovering preferences from users' historical activities. Noticeably, LBSNs have offered unparalleled access abundant heterogeneous relational information...

Graph Neural Networks (GNNs) have shown satisfying performance in various graph analytical problems. Hence, they become the de facto solution a variety of decision-making scenarios. However, GNNs could yield biased results against certain demographic subgroups. Some recent works empirically that structure input network is significant source bias for GNNs. Nevertheless, no studies systematically scrutinized which part leads to predictions any given node. The low transparency on how influences...

Graph Neural Networks (GNNs) are playing increasingly important roles in critical decision-making scenarios due to their exceptional performance and end-to-end design. However, concerns have been raised that GNNs could make biased decisions against underprivileged groups or individuals. To remedy this issue, researchers proposed various fairness notions including individual gives similar predictions existing methods rely on Lipschitz condition: they only optimize overall disregard equality...

Graph machine learning has gained great attention in both academia and industry recently. Most of the graph models, such as Neural Networks (GNNs), are trained over massive data. However, many real-world scenarios, hospitalization prediction healthcare systems, data is usually stored at multiple owners cannot be directly accessed by any other parties due to privacy concerns regulation restrictions. Federated Machine Learning (FGML) a promising solution tackle this challenge training models...

In recent years, neural models have been repeatedly touted to exhibit state-of-the-art performance in recommendation. Nevertheless, multiple studies revealed that the reported results of many recommendation cannot be reliably replicated. A primary reason is existing evaluations are performed under various inconsistent protocols. Correspondingly, these replicability issues make it difficult understand how much benefit we can actually gain from models. It then becomes clear a fair and...

In modern pavement management systems, roughness is an important indicator of performance, and it reflects the smoothness surface. International Roughness Index (IRI) de-facto metric to quantitatively analyze The with high IRI not only reduces lifetime vehicles, but also raises risk car accidents. Accurate prediction becomes a key task for system, helps transportation department refurbish in time. However, existing models are proposed on top small datasets, have poor performance. Besides,...

Driven by the powerful representation ability of Graph Neural Networks (GNNs), plentiful GNN models have been widely deployed in many real-world applications. Nevertheless, due to distribution disparities between different demographic groups, fairness high-stake decision-making systems is receiving increasing attention. Although lots recent works devoted improving GNNs and achieved considerable success, they all require significant architectural changes or additional loss functions requiring...

Researchers recently investigated to explain Graph Neural Networks (GNNs) on the access a task-specific GNN, which may hinder their wide applications in practice. Specifically, explanation methods are incapable of explaining pretrained GNNs whose downstream tasks usually inaccessible, not mention giving explanations for transferable knowledge GNNs. Additionally, only consider target models' output label space, coarse-grained and insufficient reflect model's internal logic. To address these...

Few-shot graph classification aims at predicting classes for graphs, given limited labeled graphs each class. To tackle the bottleneck of label scarcity, recent works propose to incorporate few-shot learning frameworks fast adaptations with graphs. Specifically, these accumulate meta-knowledge across diverse meta-training tasks, and then generalize such target task a disjoint set. However, existing methods generally ignore correlations among tasks while treating them independently....

The prevalence of natural hazards and extreme climatic events highlights the critical importance disaster recovery for communities. In this study, we examine impact Hurricane Harvey on private well water in Texas explore relationship between contamination water, stewardship behavior, demographic social capital characteristics. We develop a multi-level regression model that shows bonding linking index, median house income, owner-occupied houses' percentage, percentage households speaking only...