Xiao Ding

Xiao Ding

Auto-generated profile, based on publicly available information

Xiao Ding

Key Laboratory of Intelligent Computing and Signal Processing of Ministry of Education, School of Mathematical Science, Anhui University, Hefei, China

Research Focus & Metrics

Academic Career Length

4Years (2023-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
1 April 2024 in Chaos: An Interdisciplinary Journal of Nonlinear Science
Verified
Journal Article
Open access
PDF file
XDXiao Ding
Xiao Ding
LKLing-Wei Kong
Ling-Wei Kong
HZHai-Feng Zhang
Hai-Feng Zhang
Ying-Cheng LaiYing-Cheng Lai
Ying-Cheng Lai
Reconstructing complex networks and predicting the dynamics are particularly challenging in real-world applications because the available information and data are incomplete. We develop a unified collaborative deep-learning framework consisting of three modules: network inference, state estimation, and dynamical learning. The complete network structure is first inferred and the states of the unobserved nodes are estimated, based on which the dynamical learning module is activated to determine the dynamical evolution rules. An alternating parameter updating strategy is deployed to improve the inference and prediction accuracy. Our framework outperforms baseline methods for synthetic and empirical networks hosting a variety of dynamical processes. A reciprocity emerges between network inference and dynamical prediction: better inference of network structure improves the accuracy of dynamical prediction, and vice versa. We demonstrate the superior performance of our framework on an influenza dataset consisting of 37 US States and a PM2.5 dataset covering 184 cities in China.

18 January 2024 in IEEE Transactions on Network Science and Engineering
Verified
Journal Article
Closed access
HZHai-Feng Zhang
Hai-Feng Zhang
GRGuojing Ren
Guojing Ren
XDXiao Ding
Xiao Ding
LZLi Zhou
Li Zhou
XZXingyi Zhang
Xingyi Zhang
Network alignment aims to identify the corresponding nodes belonging to the same entity across different networks, which is a fundamental task in various applications. Existing embedding-based approaches usually involve two stages, namely embedding and matching. The embedding stage conducts network embedding on each network to capture the primary structural regularity. In the matching stage, a mapping function is built to project the learned embeddings to the same latent space. However, these approaches typically encounter two challenges: (1) the difficulty of unifying the elusive embedding spaces to the same latent space; (2) the difficulty of distinguishing the real anchor nodes from their neighbors, resulting in the confounding matching problem. To address these challenges, we present the Collaborative Cross-Network Embedding (CCNE) framework in this paper. This framework provides a collaborative and straightforward paradigm to better unify the two networks to the same latent space by preserving both intra- and inter-network structural features, without the need for a carefully designed mapping function. Meanwhile, a hard negative sampling strategy is adopted to distinguish anchor nodes from their sampled neighbors. Furthermore, an iterative CCNE is proposed to alleviate the scarcity of observed anchor links. Extensive experiments on real social networks demonstrate that the proposed collaborative framework outperforms current embedding-matching methods in terms of accuracy, robustness as well as compatibility.

9 June 2023 in IEEE Transactions on Instrumentation and Measurement
Verified
Journal Article
Closed access
XZXian-Jie Zhang
Xian-Jie Zhang
XDXiao Ding
Xiao Ding
HZHai-Feng Zhang
Hai-Feng Zhang
PDPan Donghui
Pan Donghui
KZKai Zhong
Kai Zhong
Due to the continuous technological innovation in industrial processes, many deep learning based methods have shown powerful capability in handing equipment status monitoring, but most of them ignore the temporal features and the dynamic changes of diverse spatial structure of the raw data. Meanwhile, these methods usually focus on handling a single downstream task, but rarely consider different tasks simultaneously. To solve these issues, this paper proposes a more flexible monitoring framework based on dynamic-multilayer graph convolution network, which can be adapted to different downstream tasks simultaneously by agile combinations of the modules according to the different industrial scenarios. First, the time series of fault samples are segmented and constructed into a feature matrix to extract the temporal information by the temporal module. At the same time, in order to fully characterize the dynamic change of different spatial structures among samples, the dynamic graph of each moment is expanded into a multilayer graph through various composition indexes. Then, the spatial information is extracted by the intra-layer and inter-layer convolution operations in the spatial module, and the fused features are applied to different downstream tasks. Finally, the experiments are performed in two different downstream tasks, namely three datasets for fault diagnosis and one dataset for remaining useful life prediction. The results of both diagnosis and prediction are better than those of the comparison algorithms, so the effectiveness and flexibility of the proposed framework are proved.

1 November 2024 in Chaos: An Interdisciplinary Journal of Nonlinear Science
Verified
Journal Article
Closed access
PDF file
HZHai-Feng Zhang
Hai-Feng Zhang
XLXin-Long Lu
Xin-Long Lu
XDXiao Ding
Xiao Ding
XZXiaoming Zhang
Xiaoming Zhang
Power flow calculation plays a significant role in the operation and planning of modern power systems. Traditional numerical calculation methods have good interpretability but high time complexity. They are unable to cope with increasing amounts of data in power systems; therefore, many machine learning based methods have been proposed for more efficient power flow calculation. Despite the good performance of these methods in terms of computation speed, they often overlook the importance of transmission lines and do not fully consider the physical mechanisms in the power systems, thereby weakening the prediction accuracy of power flow. Given the importance of the transmission lines as well as to comprehensively consider their mutual influence, we shift our focus from bus adjacency relationships to transmission line adjacency relationships and propose a physics-informed line graph neural network framework. This framework propagates information between buses and transmission lines by introducing the concepts of the incidence matrix and the line graph matrix. Based on the mechanics of the power flow equations, we further design a loss function by integrating physical information to ensure that the output results of the model satisfy the laws of physics and have better interpretability. Experimental results on different power grid datasets and different scenarios demonstrate the accuracy of our proposed model.

3 March 2025 in IEEE Transactions on Network Science and Engineering
Verified
Journal Article
Closed access
JZJia-Le Zhao
Jia-Le Zhao
XZXian-Jie Zhang
Xian-Jie Zhang
XDXiao Ding
Xiao Ding
XZXingyi Zhang
Xingyi Zhang
HZHai-Feng Zhang
Hai-Feng Zhang
Graph neural networks (GNNs), as a cutting-edge technology in deep learning, perform particularly well in various tasks that process graph structure data. However, their foundation on pairwise graphs often limits their capacity to capture latent higher-order topological semantic information. Thus, it is crucial to find a way to extract the latent higher-order information of graphs without missing the lower-order information of the original graph. To address this issue, we here develop a method to construct hypergraph based on motifs, and then a novel neural network framework, named MD-HGNN, is proposed for enhanced graph learning. Specifically, we first utilize motifs of the original graph to construct the hypergraph and eliminate nested structures within the hypergraph to prevent information redundancy. Subsequently, GNNs and hypergraph neural networks (HGNNs) are employed separately to extract the lower-order and higher-order topological semantic information of the graph. Finally, the lower-order and higher-order information are integrated to obtain an embedded representation of graph. Extensive experimental results demonstrate that MD-HGNN preserves the original lower-order graph structure information while effectively extracting higher-order features. Moreover, its performance and robustness are validated across different downstream tasks.

21 March 2024 in IEEE Transactions on Computational Social Systems
Verified
Journal Article
Closed access
KZKe-Jia Zhang
Ke-Jia Zhang
XDXiao Ding
Xiao Ding
BXBing-Bing Xiang
Bing-Bing Xiang
HZHai-Feng Zhang
Hai-Feng Zhang
ZBZhong-Kui Bao
Zhong-Kui Bao
Graph neural networks (GNNs) are efficient techniques for learning graph representations and have shown remarkable success in tackling diverse graph-related tasks. However, in the context of the neighborhood aggregation paradigm, conventional GNNs have limited capabilities in capturing the higher order structures and topological semantics of graphs. Researchers have attempted to overcome this limitation by designing new GNNs that explore the impacts of motifs to capture potentially higher order graph information. However, existing motif-based GNNs often ignore lower order connectivity patterns such as nodes and edges, which leads to poor representation of sparse networks. To address these limitations, we propose an innovative approach. First, we design convolution kernels on both motif-based and simple graphs. Second, we introduce a multilevel graph convolution framework for extracting higher order topological semantics of graphs. Our approach overcomes the limitations of prior methods, demonstrating state-of-the-art performance in downstream tasks with excellent scalability. Extensive experiments on real-world datasets validate the effectiveness of our proposed method.