Article
Computer Science, Artificial Intelligence
Thakare Kamalakar Vijay, Nitin Sharma, Debi Prosad Dogra, Heeseung Choi, Ig-Jae Kim
Summary: This paper proposes a method for abnormal event detection using spatio-temporal deep feature extractors and multiple instance learning classifier. By injecting temporal information in feature extraction, the accuracy of anomaly detection is improved. Experimental results show significant improvements in detecting long-duration anomalies and increasing detection accuracy across various categories.
EXPERT SYSTEMS WITH APPLICATIONS
(2022)
Article
Computer Science, Information Systems
Boshra Pishgoo, Ahmad Akbari Azirani, Bijan Raahemi
Summary: In this paper, a Hybrid Distributed Batch-Stream (HDBS) architecture is proposed for anomaly detection in real-time data, which combines the accuracy of batch processing with the speed and real-time features of stream processing. The combination of batch and stream processing units results in complementary models and efficient processing for real-time anomaly detection.
INFORMATION SCIENCES
(2021)
Article
Computer Science, Artificial Intelligence
Bakhtiar Amen, Syahirul Faiz, Thanh-Toan Do
Summary: This study utilizes Twitter data to detect anomalous events associated with COVID-19. By employing a distributed Directed Acyclic Graph topology framework and a lightweight algorithm, the study aggregates and processes large-scale real-time tweets, identifies, clusters, and visualizes important keywords. It was found that on a specific date, the anomaly on Twitter was prominent due to tweets mentioning casualties' updates and discussions on the pandemic.
PATTERN RECOGNITION
(2022)
Article
Computer Science, Information Systems
Wang Xiaolan, Md Manjur Ahmed, Mohd Nizam Husen, Zhao Qian, Samir Brahim Belhaouari
Summary: Network security remains a concern and anomaly-based methods have gained attention for their ability to detect novel attacks. However, building an accurate network normal pattern is challenging due to the continuous evolution and change of network data. This study presents an evolving anomaly detection method for network streaming data, which achieves high detection rates and low computational cost compared to existing methods.
INFORMATION SCIENCES
(2022)
Article
Chemistry, Multidisciplinary
Kejing Xiao, Zhaopeng Qian, Biao Qin
Summary: This paper comprehensively reviews the existing research on event detection and evolution, focusing on multi-modality event detection and event evolution under multi-modality data. The paper discusses the techniques of data representation for event detection and reviews the results of several public datasets.
APPLIED SCIENCES-BASEL
(2022)
Article
Computer Science, Interdisciplinary Applications
Renfang Wang, Hong Qiu, Xu Cheng, Xiufeng Liu
Summary: This paper proposes a flexible stream processing framework for online anomaly detection in IIoT applications. The framework utilizes a distributed computing architecture based on Docker containers to improve flexibility, migration capability, and customization. It also uses a central mediator to coordinate data stream processing tasks on different Docker nodes. Furthermore, a prediction-based online anomaly detection model is developed, which combines batch model training and data stream anomaly detection processes using LSTM neural networks and a dynamic sliding window method. Evaluation results show that the proposed framework and model achieve high accuracy and low latency, and outperform existing methods in scalability, efficiency, and adaptability for IIoT applications.
JOURNAL OF INDUSTRIAL INFORMATION INTEGRATION
(2023)
Article
Computer Science, Information Systems
Ashwin Raut, Anubhav Shivhare, Vijay Kumar Chaurasiya, Manish Kumar
Summary: The emergence of the Internet of Things is expected to produce a large amount of data streams for sensing real-world phenomena. Valuable information should be extracted in real-time to find events of interest. Traditional methods have limitations when dealing with real-time data streams. Therefore, this study proposes an adaptive clustering-based Event Detection Scheme for IoT (AEDS-IoT) to infer events of interest from data distribution patterns in streaming data.
INTERNET OF THINGS
(2023)
Article
Computer Science, Artificial Intelligence
Hao Peng, Ruitong Zhang, Shaoning Li, Yuwei Cao, Shirui Pan, Philip S. Yu
Summary: This paper presents a novel reinforced, incremental, and cross-lingual social event detection architecture, FinEvent, which models social messages into heterogeneous graphs and uses reinforcement learning algorithm to select optimal aggregation thresholds. It addresses the challenges of ambiguous event features, dispersive text contents, and multiple languages in existing event detection methods for streaming social messages, thereby improving accuracy and generalization ability.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2023)
Article
Thermodynamics
Junfeng Wu, Li Yao, Bin Liu, Zheyuan Ding, Lei Zhang
Summary: This study introduces a multi-task learning based encoder-decoder system capable of simultaneously detecting anomalies, diagnosing anomalies, and detecting events, effectively reducing monitoring burden and minimizing the risk of system faults. The use of feature matrix enables point-wise anomaly detection in an end-to-end manner, and multi-task learning is utilized to share feature matrix for anomaly diagnosis and event detection, providing crucial information for system monitoring.
ADVANCES IN MECHANICAL ENGINEERING
(2021)
Article
Computer Science, Theory & Methods
Waseem Ullah, Amin Ullah, Tanveer Hussain, Khan Muhammad, Ali Asghar Heidari, Javier Del Ser, Sung Wook Baik, Victor Hugo C. De Albuquerque
Summary: This work presents an efficient and robust framework for recognizing anomalies in surveillance Big Video Data (BVD) using Artificial Intelligence of Things (AIoT). The framework consists of two phases: instant anomaly detection and detailed anomaly analysis. A two-stream neural network is proposed to model and classify the anomalies. Extensive experiments show that the framework outperforms state-of-the-art methods in terms of accuracy.
FUTURE GENERATION COMPUTER SYSTEMS-THE INTERNATIONAL JOURNAL OF ESCIENCE
(2022)
Article
Computer Science, Artificial Intelligence
Anderson Almeida Firmino, Claudio de Souza Baptista, Anselmo Cardoso de Paiva
Summary: The growth of social media has brought both benefits and challenges, including the proliferation of hate speech. This study proposes a novel method for detecting hate speech in texts using Cross-Lingual Learning. Experimental results show that the proposed method is promising.
EXPERT SYSTEMS WITH APPLICATIONS
(2024)
Article
Computer Science, Information Systems
Chen Yang, Zhihui Du, Xiaofeng Meng, Xukang Zhang, Xinli Hao, David A. Bader
Summary: This article presents a method for detecting anomalies with high accuracy and real time from large amounts of streaming data, specifically focusing on catalog streams. The anomaly detection in catalog streams is formulated as a constrained optimization problem based on a catalog stream matrix. A novel anomaly detection algorithm, FIAD, is proposed, which includes strategies for true event identification and false alarm filtering. Different attention windows are developed to provide corresponding data for various algorithm components. The proposed method achieves a low false-positive rate of 0.04%, reduces false alarms by 98.6% compared to existing methods, and has a latency of 2.1 seconds to handle each catalog.
IEEE TRANSACTIONS ON BIG DATA
(2023)
Article
Computer Science, Artificial Intelligence
Shubham Gupta, Suman Kundu
Summary: The paper proposes a methodology named CommunityINDICATOR to find localized micro-events from social network streams. By incorporating the concept of 'separation of concerns' from software design principle, the method significantly reduces execution time compared to existing state-of-the-art methods. The method generates an interaction graph from the social stream and applies community detection and clustering algorithms to detect micro-level events. Experimental results show that CommunityINDICATOR provides up to 30% higher accuracy than EvenTweet and SEDTWik in similar execution times, and improves execution time by 11% to 51% and 17% to 57% compared to TwiiterNews and EventX algorithms, respectively, for different datasets.
EXPERT SYSTEMS WITH APPLICATIONS
(2023)
Article
Computer Science, Information Systems
Ahmed Awad, Riccardo Tommasini, Samuele Langhi, Mahmoud Kamel, Emanuele Della Valle, Sherif Sakr
Summary: Modern Big Stream Processing Solutions are evolving towards being the ultimate framework for streaming analytics by offering SQL extensions with stream-oriented primitives like windowing and Complex Event Processing. However, the point-based time semantics do not meet the requirements of many use-cases, hence the need for novel abstract operators to analyze user-defined event intervals.
INFORMATION SYSTEMS
(2022)
Article
Computer Science, Information Systems
Tajinder Singh, Madhu Kumari, Daya Sagar Gupta
Summary: Controlling data in online social networks is always a major challenge. By extracting information from various social platforms and classifying it based on similarities, users can gain direct insights into enduring topics. To address the issues of event detection and classification, this research utilizes a fast and robust method, the Siamese network, to compare social data streams and detect novel events through clustering.
CLUSTER COMPUTING-THE JOURNAL OF NETWORKS SOFTWARE TOOLS AND APPLICATIONS
(2022)
Article
Computer Science, Artificial Intelligence
Xusheng Zhao, Qiong Dai, Jia Wu, Hao Peng, Mingsheng Liu, Xu Bai, Jianlong Tan, Senzhang Wang, Philip S. Yu
Summary: In this paper, a novel Reinforced Tensor Graph Neural Network (RTGNN) framework is proposed to effectively learn multi-view graph representation by reinforcing inter- and intra-graph aggregation. The framework utilizes tensor decomposition to extract the graph structure features (GSFs) of each view, which contain the correlation information among multiple views. Furthermore, a reinforcement learning guided scheme is developed to automatically calculate the optimal filtering threshold for each view, improving the performance of intra-graph aggregation.
IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING
(2023)
Article
Computer Science, Artificial Intelligence
Hao Peng, Ruitong Zhang, Shaoning Li, Yuwei Cao, Shirui Pan, Philip S. Yu
Summary: This paper presents a novel reinforced, incremental, and cross-lingual social event detection architecture, FinEvent, which models social messages into heterogeneous graphs and uses reinforcement learning algorithm to select optimal aggregation thresholds. It addresses the challenges of ambiguous event features, dispersive text contents, and multiple languages in existing event detection methods for streaming social messages, thereby improving accuracy and generalization ability.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2023)
Article
Biochemistry & Molecular Biology
Juxu Li, Man Li, Weimin Wang, Dong Wang, Yuwei Hu, Yunyun Zhang, Xiaoquan Zhang
Summary: In this study, the morphological and physiological mechanisms of cytoplasmic inheritance stigma exsertion trait expression in a tobacco line (MSK326SE) were investigated. It was found that the exserted stigma phenotype of MSK326SE was mainly caused by corolla shortening, which was stable under different environmental temperature. The difference in cell division and expansion in different flower organs, as well as the decrease in corolla epidermal cell size, were the main factors contributing to the stigma exsertion of MSK326SE. Exogenous JA could further enhance the stigma exsertion degree.
Article
Chemistry, Multidisciplinary
Hao Peng, Yulin Xu, Changjiang Zhou, Ranran Pei, Jingsheng Miao, He Liu, Chuluo Yang
Summary: This study develops three thermally activated delayed fluorescent (TADF) emitters based on fluorene-spiro-acridine derivatives as donors, which exhibit high external quantum efficiency (EQE) and low efficiency roll-off. The emitters achieve high photoluminescence quantum yields (PLQYs) of 96-99% due to the rigid donor and acceptor composition. The enhancement of reverse intersystem crossing (k(RISC)) in IA-TRZ and the acceleration of radiative decay rate (k(r)) in 2S-TRZ and IT-TRZ contribute to their high electroluminescent performance.
ADVANCED FUNCTIONAL MATERIALS
(2023)
Article
Chemistry, Multidisciplinary
Ranran Pei, Yulin Xu, Jingsheng Miao, Hao Peng, Zhanxiang Chen, Changjiang Zhou, He Liu, Chuluo Yang
Summary: Simultaneously achieving high efficiency and low efficiency roll-off is crucial for the further application of thermally activated delayed fluorescent (TADF) emitters. This study proposes a quinolino-acridine (QAc) donor, composed of two acridine units with planar (pAc) and bended (bAc) geometries, combined with triazine to assemble a TADF emitter QAc-TRZ. The pAc ensures decent TADF behavior through interaction with triazine, while the bAc enhances radiative decay through delocalization of the highest occupied molecular orbital (HOMO). Notably, QAc-TRZ enables a highly efficient organic light emitting diode (OLED) with a maximum external quantum efficiency (EQE) of 37.3%, and maintains efficiencies of 36.3% and 31.7% under 100/1000 cd m(-2), respectively, even at 10,000 cd m(-2).
ANGEWANDTE CHEMIE-INTERNATIONAL EDITION
(2023)
Article
Thermodynamics
Hao Peng, Baofeng Wang, Wenxiu Li, Fengling Yang, Fangqin Cheng
Summary: Coal slime added during coal gangue combustion affects combustion characteristics and leads to NO emissions, causing environmental pollution. Reducing NO emissions is possible in O-2/CO2 atmospheres. This study investigates the co-combustion of coal gangue and coal slime in O-2/CO2 atmospheres, finding that coal slime enhances combustion activity and co-combustion shows synergistic effects. The proportion of pyrrole and nitrogen oxide in the ashes increases with temperature, while pyridine and quaternary nitrogen disappear, and total nitrogen content in ash decreases. These findings contribute to understanding the co-combustion process and provide data for practical industrial applications.
JOURNAL OF THERMAL SCIENCE
(2023)
Article
Computer Science, Artificial Intelligence
Xingcheng Fu, Jianxin Li, Jia Wu, Jiawen Qin, Qingyun Sun, Cheng Ji, Senzhang Wang, Hao Peng, Philip S. Yu
Summary: Current graph neural networks (GNNs) based on Euclidean space embedding struggle to represent diverse graph geometric properties. To address this issue, we propose the Adaptive Curvature Exploration Geometric Graph Neural Network, which automatically learns high-quality graph representations and explores embedding spaces with optimal curvature.
KNOWLEDGE AND INFORMATION SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Tao Zhang, Congying Xia, Zhiwei Liu, Shu Zhao, Hao Peng, Philip Yu
Summary: This article proposes an adversarial adaptive augmentation method to address the problems in cross-domain named entity recognition. It integrates adversarial strategy into a multi-task learner to improve the quality of domain adaptive data generation and extract domain-invariant features to solve the domain shift and label-sparsity problems. The article also introduces a progressive domain-invariant feature distillation framework that uses multi-grained Maximum Mean Discrepancy approach to extract multi-level domain invariant features for knowledge transfer across domains. Extensive comparative experiments on four English and two Chinese benchmarks demonstrate the importance of adversarial augmentation and effective adaptation from high-resource domains to low-resource target domains. Comparison with two vanilla and four latest baselines shows state-of-the-art performance and superiority in both zero-resource and minimal-resource scenarios.
ACM TRANSACTIONS ON ASIAN AND LOW-RESOURCE LANGUAGE INFORMATION PROCESSING
(2023)
Article
Biodiversity Conservation
Achyut Kumar Banerjee, Jiakai Wang, Hui Feng, Yuting Lin, Xinru Liang, Minghui Yin, Hao Peng, Weixi Li, Tengjiao Li, Wuxia Guo, Yelin Huang
Summary: Understanding genetic patterns and environmental influence on Thespesia populnea is crucial for its conservation and sustainable use. This study focused on the species' phylogeographic pattern, genetic diversity distribution, and range dynamics under changing environmental conditions. Findings suggested sustainable use of genetically diverse populations, measures for preserving lineages' evolutionary potential, and prioritized conservation areas.
BIODIVERSITY AND CONSERVATION
(2023)
Article
Computer Science, Artificial Intelligence
Yuecen Wei, Xingcheng Fu, Dongqi Yan, Qingyun Sun, Hao Peng, Jia Wu, Jinyan Wang, Xianxian Li
Summary: Most social networks can be modeled as heterogeneous graphs, but existing privacy-preserving methods neglect the significance of high-order semantic information. To address this issue, we propose HeteSDG, a novel heterogeneous graph neural network that provides double privacy guarantee and performance trade-off. We reveal the privacy leakage in heterogeneous graphs and design a two-stage mechanism based on differential privacy to defend against privacy attacks and provide stronger interpretation.
KNOWLEDGE AND INFORMATION SYSTEMS
(2023)
Article
Chemistry, Physical
Jinghua Li, Hao Peng, Bing Luo, Jiamei Cao, Lijing Ma, Dengwei Jing
Summary: A new heterostructure between Ti3C2 MXene quantum dot and 3D macroscopic porous graphitic carbon nitride (PGCN) was successfully obtained using in situ electrostatic self-assembly techniques. The optimized Ti3C2 QD/PGCN composites exhibit significantly enhanced photocatalytic H2 evolution rate compared to pristine CN and PGCN. The improved performance is attributed to the synergistic effects of Ti3C2 quantum dots and the good photothermal conversion ability of PGCN.
JOURNAL OF COLLOID AND INTERFACE SCIENCE
(2023)
Article
Computer Science, Artificial Intelligence
Jianxin Li, Hao Peng, Yuwei Cao, Yingtong Dou, Hekai Zhang, Philip S. Yu, Lifang He
Summary: Graph neural networks (GNNs) are widely used in deep learning for graph analysis tasks. However, current methods ignore heterogeneity in real-world graphs and fail to capture content-based correlations between nodes. In this paper, we propose a novel HAE framework and a HAE(GNN) model that incorporates meta-paths and meta-graphs for rich, heterogeneous semantics and leverages self-attention mechanism for exploring content-based interactions between nodes.
IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING
(2023)
Article
Computer Science, Artificial Intelligence
Hao Peng, Jianxin Li, Zheng Wang, Renyu Yang, Mingsheng Liu, Mingming Zhang, Philip S. Yu, Lifang He
Summary: Luce is a life-long predictive model that addresses the lack of recent sold prices and sparsity of house data in property valuation. It utilizes a heterogeneous information network (HIN) to organize house data and employs a Graph Convolutional Network (GCN) and Long Short Term Memory (LSTM) network to extract spatial and temporal information for accurate valuation.
IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING
(2023)
Article
Engineering, Electrical & Electronic
Jun Dong, Fan Yin, Taixing Jiang, Xiang Zhong, Yang Yang, Hao Peng
Summary: This work presents and studies a low-cost and mechanical reconfigurable substrate integrated waveguide (SIW) equalizer. Indium tin oxides (ITO) are introduced instead of Tantalum Nitride (TaN) or absorbing material as the resistive element in SIW equalizer, providing more structural stability, excellent high frequency characteristic, and easy integration with traditional printed circuit boards. The proposed equalizer based on ITO Conductive Film demonstrates mechanical reconfigurability, low cost, and high stability, making ITO a good candidate for high millimeter-wave band equalizer design.
PROGRESS IN ELECTROMAGNETICS RESEARCH LETTERS
(2023)
Article
Chemistry, Physical
Hao Peng, Zichen Di, Pan Gong, Fengling Yang, Fangqin Cheng
Summary: A novel double chemical looping process is proposed for the co-production of high purity hydrogen and carbon monoxide. The system achieves high efficiency and economic competitiveness, making it a promising method for hydrogen production.
GREEN ENERGY & ENVIRONMENT
(2023)
