Author name cluster

Qing He 0003

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

5 papers

1 author row

ICLR Conference 2024 Conference Paper

Boosting the Adversarial Robustness of Graph Neural Networks: An OOD Perspective

Kuan Li
Yiwen Chen
Yang Liu 0200
Jin Wang 0007
Qing He 0003
Minhao Cheng
Xiang Ao 0001

Current defenses against graph attacks often rely on certain properties to eliminate structural perturbations by identifying adversarial edges from normal edges. However, this dependence makes defenses vulnerable to adaptive (white-box) attacks from adversaries with the same knowledge. Adversarial training seems to be a feasible way to enhance robustness without reliance on artificially designed properties. However, in this paper, we show that it can lead to models learning incorrect information. To solve this issue, we re-examine graph attacks from the out-of-distribution (OOD) perspective for poisoning and evasion attacks and introduce a novel adversarial training paradigm incorporating OOD detection. This approach strengthens the robustness of Graph Neural Networks (GNNs) without reliance on prior knowledge. To further evaluate adaptive robustness, we develop adaptive attacks against our methods, revealing a trade-off between graph attack efficacy and defensibility. Through extensive experiments over 25,000 perturbed graphs, our method could still maintain good robustness against both adaptive and non-adaptive attacks. The code is provided at https://github.com/likuanppd/GOOD-AT.

Details

ICLR Conference 2023 Conference Paper

Revisiting Graph Adversarial Attack and Defense From a Data Distribution Perspective

Kuan Li
Yang Liu 0200
Xiang Ao 0001
Qing He 0003

Recent studies have shown that structural perturbations are significantly effective in degrading the accuracy of Graph Neural Networks (GNNs) in the semi-supervised node classification (SSNC) task. However, why the gradient-based methods are so destructive is rarely explored. In this work, we discover an interesting phenomenon: the adversarial edges are not uniformly distributed on the graph. Nearly all perturbations are generated around the training nodes in poisoning attack. Combined with this phenomenon, we provide an explanation for the effectiveness of the gradient-based attack method from a data distribution perspective and revisit both poisoning attack and evasion attack in SSNC. From this new perspective, we empirically and theoretically discuss some other attack tendencies. Based on the analysis, we provide nine practical tips on both attack and defense and meanwhile leverage them to improve existing attack and defense methods. Moreover, we design a fast attack method and a self-training defense method, which outperform the state-of-the-art methods and can effectively scale to large graphs like ogbn-arxiv. We conduct extensive experiments on four benchmark datasets to verify our claims.

Details

ECAI Conference 2020 Conference Paper

Capturing Attraction Distribution: Sequential Attentive Network for Dwell Time Prediction

Tianxin Wang
Jingwu Chen
Fuzhen Zhuang
Leyu Lin
Feng Xia 0006
Lihuan Du
Qing He 0003

In article recommendation, the dwell time is an important metric to measure user engagement on content and has been widely used as a proxy for user satisfaction. Therefore, predicting the dwell time is very helpful for making better recommendations and improving user experience. Modeling the interaction between user and content is the key for dwell time prediction. However, conventional methods usually model the content with document-level representation in a non-personalized way, which ignores the natural reading process of the reader and the reader attraction in sub-document level, this might lead to a bias for analyzing the user reading behavior. Since the attraction level of different parts is different for the user, the user attention changes dynamically while reading. The former content affects the reading for the latter content via the change of attraction level. Therefore, considering the attraction level of each part of the article, i. e. , attraction distribution, is quite necessary for content modeling. In this paper, we propose the Sequential Attentive Network (SAN) for dwell time prediction, which effectively models the attraction distribution of the article reading process. We collect the data from WeChat, a widely-used mobile app in China, for experiments. The results demonstrate the advantages of our model over several competitive baselines on dwell time prediction.

Details

UAI Conference 2016 Conference Paper

Online Bayesian Multiple Kernel Bipartite Ranking

Changying Du
Changde Du
Guoping Long
Qing He 0003
Yucheng Li 0002

Bipartite ranking aims to maximize the area under the ROC curve (AUC) of a decision function. To tackle this problem when the data appears sequentially, existing online AUC maximization methods focus on seeking a point estimate of the decision function in a linear or predefined single kernel space, and cannot learn effective kernels automatically from the streaming data. In this paper, we first develop a Bayesian multiple kernel bipartite ranking model, which circumvents the kernel selection problem by estimating a posterior distribution over the model weights. To make our model applicable to streaming data, we then present a kernelized online Bayesian passive-aggressive learning framework by maintaining a variational approximation to the posterior based on data augmentation. Furthermore, to efficiently deal with large-scale data, we design a fixed budget strategy which can effectively control online model complexity. Extensive experimental studies confirm the superiority of our Bayesian multi-kernel approach.

Details

ECAI Conference 2014 Conference Paper

Nonparametric Bayesian Multi-Task Large-margin Classification

Changying Du
Jia He 0001
Fuzhen Zhuang
Yuan (Alan) Qi
Qing He 0003

In this paper, we present a nonparametric Bayesian multi-task large-margin classification model which can cluster tasks into the most appropriate number of groups and induce flexible model sharing within each task group simultaneously. Specifically, we first show a very simple method to integrate large margin learning with hierarchical Bayesian models by employing an important variant of the standard SVMi. e. proximal SVM (PSVM)whose loss function is used to define a novel likelihood function. And then we assume that the model parameter of each task consists of two parts: one is shared within each task group (group-level parameter) while the other is specific to each distinct task (task rescaling parameter). A Dirichlet process prior is imposed on the group-level parameter while the task rescaling parameter is assigned a one-mean Laplace prior. Finally the parameter of a task is the corresponding group parameter times its specific rescaling parameter. We give efficient Markov chain Monte Calo (MCMC) algorithm to conduct model inference. Experiments on the Landmine detection data and the UCI Yeast data demonstrate the effectiveness of our method.

Details