Junchen Ye

Dr. Junchen Ye


Research Focus & Metrics

Academic Career Length

6Years (2021-Present)

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

Representative Publications

View All Publications
17 November 2021 in IEEE Transactions on Knowledge and Data Engineering
Verified
Journal Article
Closed access
BDBowen Du
Bowen Du
XSXuanxuan Sun
Xuanxuan Sun
JYJunchen Ye
Junchen Ye
KCKe Cheng
Ke Cheng
JWJingyuan Wang
Jingyuan Wang
LSLeilei Sun
Leilei Sun
Multivariate time series anomaly detection has great potentials in many practical applications. Extreme unbalanced training set and noise interference make it challenging to accurately capture the distribution of normal data and then detect anomalies. Existing AutoEncoder(AE)-based approaches are lack of effective regularization method specially designed for anomaly detection tasks thus easily overfitting while Generative Adversarial Network(GAN)-based approaches are mostly trained under the hypothesis of pollution-free training set, which means the training set is all composed of normal samples and that is hard to satisfy in practice. To tackle these problems, in this paper we propose a GAN based anomaly detection method for multivariate time series named FGANomaly (letter F is for Filter). The core idea is to filter possible anomalous samples with pseudo-labels before training the discriminator thus to capture the distribution of normal data as precise as possible. In addition, we design a novel training objective for the generator, which leads the generator to concentrate more on plausible normal data and ignore anomalies. We conducted comprehensive experiments on four public datasets, and the experimental results show the superiority of our method over baselines in both performance and robustness.

24 August 2024
Verified
Conference Paper
Open access
KCKe Cheng
Ke Cheng
LPLinzhi Peng
Linzhi Peng
PWPengyang Wang
Pengyang Wang
JYJunchen Ye
Junchen Ye
LSLeilei Sun
Leilei Sun
BDBowen Du
Bowen Du
Knowledge Tracing aims to assess student learning states by predicting their performance in answering questions. Different from the existing research which utilizes fixed-length learning sequence to obtain the student states and regards KT as a static problem, this work is motivated by three dynamical characteristics: 1) The scales of students answering records are constantly growing; 2) The semantics of time intervals between the records vary; 3) The relationships between students, questions and concepts are evolving. The three dynamical characteristics above contain the great potential to revolutionize the existing knowledge tracing methods. Along this line, we propose a Dynamic Graph-based Knowledge Tracing model, namely DyGKT. In particular, a continuous-time dynamic question-answering graph for knowledge tracing is constructed to deal with the infinitely growing answering behaviors, and it is worth mentioning that it is the first time dynamic graph learning technology is used in this field. Then, a dual time encoder is proposed to capture long-term and short-term semantics among the different time intervals. Finally, a multiset indicator is utilized to model the evolving relationships between students, questions, and concepts via the graph structural feature. Numerous experiments are conducted on five real-world datasets, and the results demonstrate the superiority of our model. All the used resources are publicly available at https://github.com/PengLinzhi/DyGKT.

24 August 2024
Verified
Conference Paper
Open access
TZTao Zou
Tao Zou
YMYuhao Mao
Yuhao Mao
JYJunchen Ye
Junchen Ye
BDBowen Du
Bowen Du
Dynamic graph learning equips the edges with time attributes and allows multiple links between two nodes, which is a crucial technology for understanding evolving data scenarios like traffic prediction and recommendation systems. Existing works obtain the evolving patterns mainly depending on the most recent neighbor sequences. However, we argue that whether two nodes will have interaction with each other in the future is highly correlated with the same interaction that happened in the past. Only considering the recent neighbors overlooks the phenomenon of repeat behavior and fails to accurately capture the temporal evolution of interactions. To fill this gap, this paper presents RepeatMixer, which considers evolving patterns of first and high-order repeat behavior in the neighbor sampling strategy and temporal information learning. Firstly, we define the first-order repeat-aware nodes of the source node as the destination nodes that have interacted historically and extend this concept to high orders as nodes in the destination node's high-order neighbors. Then, we extract neighbors of the source node that interacted before the appearance of repeat-aware nodes with a slide window strategy as its neighbor sequence. Next, we leverage both the first and high-order neighbor sequences of source and destination nodes to learn temporal patterns of interactions via an MLP-based encoder. Furthermore, considering the varying temporal patterns on different orders, we introduce a time-aware aggregation mechanism that adaptively aggregates the temporal representations from different orders based on the significance of their interaction time sequences. Experimental results demonstrate the superiority of RepeatMixer over state-of-the-art models in link prediction tasks, underscoring the effectiveness of the proposed repeat-aware neighbor sampling strategy.

24 August 2024
Verified
Conference Paper
Open access
KCKe Cheng
Ke Cheng
PLPeng Linzhi
Peng Linzhi
JYJunchen Ye
Junchen Ye
LSLeilei Sun
Leilei Sun
BDBowen Du
Bowen Du
Structure encoding has proven to be the key feature to distinguishing links in a graph. However, Structure encoding in the temporal graph keeps changing as the graph evolves, repeatedly computing such features can be time-consuming due to the high-order subgraph construction. We develop the Co-Neighbor Encoding Schema (CNES) to address this issue. Instead of recomputing the feature by the link, CNES stores information in the memory to avoid redundant calculations. Besides, unlike the existing memory-based dynamic graph learning method that stores node hidden states, we introduce a hashtable-based memory to compress the adjacency matrix for efficient structure feature construction and updating with vector computation in parallel. Furthermore, CNES introduces a Temporal-Diverse Memory to generate long-term and short-term structure encoding for neighbors with different structural information. A dynamic graph learning framework, Co-Neighbor Encoding Network (CNE-N), is proposed using the aforementioned techniques. Extensive experiments on thirteen public datasets verify the effectiveness and efficiency of the proposed method.

7 March 2023 in Georisk: Assessment and Management of Risk for Engineered Systems and Geohazards
Verified
Journal Article
Closed access
BDBowen Du
Bowen Du
TZTao Zou
Tao Zou
JYJunchen Ye
Junchen Ye
XTXuyan Tan
Xuyan Tan
KCKe Cheng
Ke Cheng
WCWeizhong Chen
Weizhong Chen
Accurate prediction of the future mechanical behaviour of the underground structure is important for the traffic system. However, the existing works mostly just predicted one type of property and failed to study the influence of different properties in the tunnel. Besides, most of them predicted future behaviours without considering the external influence of the environment like temperature and water pressure. In this paper, we propose a multi-task prediction model named MSTNet which combines different types of indicators and external factors for capturing the temporal and spatial characteristics in the tunnel. Firstly, we integrate time series of multiple indicators and build a deep learning algorithm based on a graph neural network and recurrent network to capture the temporal, spatial and external impacts in the tunnel. Then, we have a case study on the Wuhan Yangtze River tunnel which contains the comparison of different components in our model, the compared results between the single indicator and multiple indicators and the performance analysis among traditional models. From the experiment results, we could find that the ability of the MSTNet model is superior to other methods, whose capability achieved over 93% and 94% in strain variation and joint opening sensors on Pearson Correlation Coefficient (PCC) in the next 45 days, respectively.

1 January 2025 in Structural Control and Health Monitoring
Verified
Journal Article
Open access
XTXuyan Tan
Xuyan Tan
WCWeizhong Chen
Weizhong Chen
LFLixiang Fan
Lixiang Fan
JYJunchen Ye
Junchen Ye
BDBowen Du
Bowen Du
Research on early warning of tunnel anomalies is fundamental for achieving intelligent management. However, most current methods for determining early‐warning values of tunnel mechanics indicators are different to couple the nonlinear variation property and spatial positional difference. Therefore, this research presents a novel approach to tunnel early warning based on deep autoregressive learning method (DL‐AR) that considers the spatiotemporal correlations of structural mechanics responses, specifically tailored to dynamically determine the warning thresholds at different spatial positions. The methodology introduces a framework for predictive modeling and instantiates it on a typical underwater shield tunnel. After thoroughly learning the temporal and spatial correlations of structural mechanical responses, accurate predictions are made for the evolving trends of structural behaviors and the probabilities to the reasonable fluctuation range. Based on these predictions, spatially varying alert thresholds for structural behaviors are proposed. To ensure the reliability of the proposed model, a series of discussions and validation experiments are conducted. Results indicate that the proposed model effectively captured the spatiotemporal characteristics of structural evolution and identified alert ranges, defining permissible variations in structural trends. The prediction results showed near to 99% consistency with actual data, a 5% enhancement compared to classical models. Any deviation beyond this range triggers an early warning, demonstrating the efficacy of model in anticipating and responding to potential structural issues.