Article
Computer Science, Artificial Intelligence
Botao Jiao, Yinan Guo, Dunwei Gong, Qiuju Chen
Summary: This study proposes a dynamic ensemble selection method to deal with concept drift in imbalanced data streams. By using a novel technique to generate new instances and selecting the optimal combination based on candidate classifier performance, the proposed method outperforms others in terms of classification accuracy and tracking new concepts.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2022)
Article
Computer Science, Artificial Intelligence
Hang Yu, Weixu Liu, Jie Lu, Yimin Wen, Xiangfeng Luo, Guangquan Zhang
Summary: This paper focuses on concept drift across multiple data streams, particularly in situations where the drift of each data stream cannot be detected in time due to slight underlying distribution drifts. The authors propose a method that constructs a distribution free test statistic and designs an online learning algorithm to detect concept drift.
PATTERN RECOGNITION
(2023)
Article
Computer Science, Artificial Intelligence
Mingyuan Wang, Adrian Barbu
Summary: This paper introduces online feature selection methods and compares them with traditional screening methods. The experiments show that online screening methods can handle modern datasets with streaming input, sparsity, and concept drift, and generate the same feature importance as their offline versions with faster speed and less storage requirements.
IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING
(2023)
Review
Computer Science, Artificial Intelligence
Andres L. Suarez-Cetrulo, David Quintana, Alejandro Cervantes
Summary: This survey reviews the methods and research trends for dealing with concept drift in continuous data streams. It introduces the field of data stream learning, discusses mechanisms for adapting to or detecting concept drifts, and presents supervised and non-supervised methods for handling seasonality in data streams. The aim is to provide future research directions in handling shifts and recurrences in continuous learning scenarios.
EXPERT SYSTEMS WITH APPLICATIONS
(2023)
Article
Computer Science, Artificial Intelligence
Sanmin Liu, Shan Xue, Jia Wu, Chuan Zhou, Jian Yang, Zhao Li, Jie Cao
Summary: This article introduces an active learning framework based on a dual-query strategy and Ebbinghaus's law of human memory cognition to address the common problems in classification methods for streaming data. The framework, called CogDQS, significantly reduces the cost of labeling by sampling representative instances for manual annotation and determines drift from noise based on the Ebbinghaus forgetting curve. Simulations show that CogDQS produces accurate, stable models with good generalization ability at minimal labeling, storage, and computation costs.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Automation & Control Systems
Cobbinah B. Mawuli, Liwei Che, Jay Kumar, Salah Ud Din, Zhili Qin, Qinli Yang, Junming Shao
Summary: Distributed data stream mining is gaining attention due to the collection of huge amounts of streaming data from different locations. However, privacy concerns have received little investigation in existing studies. This article presents FedStream, a federated learning framework that captures evolving concepts on distributed concept-drifting data streams while preserving privacy among participating clients. Extensive experiments demonstrate the superiority of FedStream over state-of-the-art methods.
IEEE TRANSACTIONS ON SYSTEMS MAN CYBERNETICS-SYSTEMS
(2023)
Article
Computer Science, Information Systems
Conor Fahy, Shengxiang Yang
Summary: The proposed MDSC algorithm addresses the challenges of change in dynamic stream mining by using multiple density clustering and outlier buffering. Experimental results demonstrate its superior performance on a variety of real and synthetic data streams, showing good scalability and noise robustness.
IEEE TRANSACTIONS ON BIG DATA
(2022)
Article
Computer Science, Artificial Intelligence
Weike Liu, Hang Zhang, Zhaoyun Ding, Qingbao Liu, Cheng Zhu
Summary: The paper proposed a comprehensive active learning method (CALMID) for handling multiclass imbalanced streaming data with concept drift. Novel uncertainty strategies and sample weight formulas were designed, and experimental results showed that CALMID outperformed existing algorithms in various imbalance and drift scenarios.
KNOWLEDGE-BASED SYSTEMS
(2021)
Article
Computer Science, Artificial Intelligence
Hang Zhang, Weike Liu, Qingbao Liu
Summary: This paper proposes a novel approach named ROALE-DI to handle the challenges of concept drift and class imbalance. By applying a reinforcement mechanism to increase the weight of dynamic classifiers, the classification performance is improved. The hybrid labeling strategy is introduced to determine the real label of instances, reducing the labeling cost.
IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING
(2022)
Article
Computer Science, Artificial Intelligence
Si -si Zhang, Jian-wei Liu, Xin Zuo
Summary: Recent years have seen a growing interest in online incremental learning, but there are three major challenges - concept drift, catastrophic forgetting, and learning of latent representation. An Adaptive Online Incremental Learning algorithm (AOIL) is proposed to address these difficulties by utilizing auto-encoder with memory module and self-attention mechanism. Extensive experiments show that AOIL outperforms other state-of-the-art methods, demonstrating promising results.
APPLIED SOFT COMPUTING
(2021)
Article
Computer Science, Artificial Intelligence
Alberto Cano, Bartosz Krawczyk
Summary: This article introduces a novel online ensemble classifier called ROSE, which is capable of handling challenges in data streams such as concept drift and class imbalance. ROSE features online training of base classifiers, online detection of concept drift, sliding window per class to handle imbalance, and self-adjusting bagging. Experimental results demonstrate that ROSE performs well in various data stream mining tasks.
Article
Computer Science, Information Systems
Shruti Arora, Rinkle Rani, Nitin Saxena
Summary: Concept drift is a crucial issue in streaming data that requires attention in machine learning models. This paper proposes a dynamic ensemble classifier and a selective ensemble method (SETL) that can effectively detect and adapt to drifting concepts, outperforming existing algorithms in performance metrics.
CLUSTER COMPUTING-THE JOURNAL OF NETWORKS SOFTWARE TOOLS AND APPLICATIONS
(2023)
Article
Computer Science, Artificial Intelligence
Ege Berkay Gulcan, Fazli Can
Summary: Many real-world applications adopt multi-label data streams as the need for algorithms to deal with rapidly changing data increases. We propose a novel algorithm called Label Dependency Drift Detector (LD3) for unsupervised concept drift detection in multi-label data streams. Our study shows that LD3 provides better predictive performance than other detectors on both real-world and synthetic data streams.
ARTIFICIAL INTELLIGENCE REVIEW
(2023)
Article
Computer Science, Artificial Intelligence
Xiulin Zheng, Peipei Li, Xuegang Hu, Kui Yu
Summary: Mining non-stationary streams poses challenges due to their infinite length, dynamic characteristics, concept drift, concept evolution, and limited labeled data. Existing supervised methods may result in poor performance and efficiency in the presence of scarce labeled data. This paper proposes a semi-supervised framework ESCR to detect recurring concept drifts and concept evolution in data streams with partially labeled data. The framework utilizes clustering-based classifiers, Jensen-Shannon divergence for change detection, and outlier monitoring for concept evolution, while also improving efficiency through recursive function and dynamic programming. Extensive experiments show the effectiveness and efficiency of ESCR compared to other semi-supervised methods.
KNOWLEDGE-BASED SYSTEMS
(2021)
Article
Computer Science, Information Systems
Ning Liu, Jianhua Zhao
Summary: In this paper, a streaming data classification algorithm based on hierarchical concept drift and online ensemble (SCHCDOE) is proposed to improve the performance of online learning in real-time distribution of streaming data. By utilizing newly arrived data instances and an anomaly detection mechanism, the algorithm achieves quick response to concept drift and efficient updates of the learning model, showing good performance in experiments.
Article
Computer Science, Artificial Intelligence
Hongwei Yue, Yufeng Huang, Chi-Man Vong, Yingying Jin, Zhiqiang Zeng, Mingqi Yu, Chuangquan Chen
Summary: Scene text recognition (STR) is widely used in industrial and commercial fields. However, existing methods struggle with processing text images that have defects such as low contrast, blur, low resolution, and insufficient illumination. To address these challenges, a novel network called NRSTRNet is proposed, which effectively reduces noise and achieves superior accuracy in text recognition.
INTERNATIONAL JOURNAL OF COMPUTATIONAL INTELLIGENCE SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Jintao Huang, Wenbin Qian, Chi-Man Vong, Weiping Ding, Wenhao Shu, Qin Huang
Summary: This paper proposes a new multi-label feature selection algorithm that effectively resolves existing algorithms' issues through three innovative procedures. First, a new similarity relation metric is proposed to deal with hybrid data types effectively. Second, a label enhancement algorithm is designed to enhance and transform the logical labels into a label distribution by fully considering the analytic hierarchy process (AHP) embedded with label correlation, which can automatically identify the significance of different labels. Third, a feature weighting evaluation is redesigned in the feature selection process to obtain the optimal feature subset through feature ranking directly. Under these proposed procedures, multi-label feature selection can effectively utilize the abundant semantic information of the label significance and can significantly improve the operating accuracy and efficiency simultaneously.
IEEE TRANSACTIONS ON EMERGING TOPICS IN COMPUTATIONAL INTELLIGENCE
(2023)
Article
Computer Science, Artificial Intelligence
Zhen Jia, Zhenbao Liu, Chi -Man Vong, Shengdong Wang, Yongyi Cai
Summary: This study uses a source circuit model with sufficient data to solve the problem of fault diagnosis in a target circuit with a lack of data. A deep transfer kernel extreme learning machine auto encoder (DKEA) model is designed, where the Gaussian error linear units (GELD) activation function is used to describe the probability of neuron input and the kernel extreme learning machine is employed as a classifier for the diagnosis task.
EXPERT SYSTEMS WITH APPLICATIONS
(2023)
Article
Computer Science, Artificial Intelligence
Jie Du, Yanhong Zhou, Peng Liu, Chi-Man Vong, Tianfu Wang
Summary: A parameter-free loss function is proposed for deep learning image classification tasks, which reduces training time, pays more attention to minority classes, and achieves higher accuracy compared to existing loss functions.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Erhao Zhou, Chi Man Vong, Yusuke Nojima, Shitong Wang
Summary: This study aims to enhance the generalization performance of the first-order TSK fuzzy rules by determining the weight of each rule and avoiding the intractable training of the consequent parts. A mathematically equivalent bridge is built between a Gaussian mixture model and a fully interpretable first-order TSK fuzzy system, resulting in a simpler expression and enhanced generalization performance. The proposed training method effectively provides an analytical solution to the weight of each rule.
IEEE TRANSACTIONS ON FUZZY SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Hongyan Li, Chi Man Vong, Zhonglin Wan
Summary: The current work presents a new multi-graph embedding collaborative disambiguation PLL algorithm (PL-MGECD) that introduces a unified framework for graph-based PLL, adopts various graph structures, and proposes an efficient optimization algorithm. Extensive experiments show that PL-MGECD has a competitive or superior performance over traditional PLL methods.
NEURAL COMPUTING & APPLICATIONS
(2023)
Article
Computer Science, Artificial Intelligence
Peng Liu, Jie Du, Chi -Man Vong
Summary: In this work, a lightweight model called SMF-Net is proposed to simultaneously alleviate the issues of over-fitting and under-fitting using a novel sequential structure of multi-scale feature learning. Compared to both deep and lightweight models, the proposed sequential structure in SMF-Net can easily extract features with larger receptive fields for multi-scale feature learning with only a few and linearly increased model parameters.
Article
Automation & Control Systems
Yichen Sun, Chi Man Vong, Shitong Wang
Summary: In this study, a fast AUC maximizing learning machine called rho-AUCCVM is proposed, which incorporates the generalized AUC metric and the core vector machine technique for simultaneous outlier detection. rho-AUCCVM has the advantages of CVM and can automatically determine the importance of the minority class or the upper bound of noises.
IEEE TRANSACTIONS ON CYBERNETICS
(2023)
Article
Nursing
Jrywan N. Huang, Margit Gerardi, Olivia Yeargain, Tracy Senterfitt, Maria Saldiva
Summary: More than half of veterans diagnosed with OUD have experienced hospitalization or death due to overdose. Telephone outreach improves access to naloxone for high-risk populations. A nurse-led intervention team successfully increased the naloxone prescription rates for at-risk veterans at the facility within three months.
ISSUES IN MENTAL HEALTH NURSING
(2023)
Article
Computer Science, Artificial Intelligence
Jintao Huang, Chi-Man Vong, Wenbin Qian, Qin Huang, Yimin Zhou
Summary: This paper proposes a novel LDL framework called OLD_RVFL+, which can effectively and accurately handle online data streams. It includes LD_RVFL+ network, a weight update module based on LD_RVFL+, and a label thresholding module for improved accuracy.
IEEE TRANSACTIONS ON EMERGING TOPICS IN COMPUTATIONAL INTELLIGENCE
(2023)
Article
Computer Science, Artificial Intelligence
Qi Lai, Chi-Man Vong, Jianhang Zhou, Yimin Zhou, C. L. Philip Chen
Summary: Multiview multi-instance multilabel learning (M3L) is a hot research topic for modeling complex real-world objects. However, existing methods suffer from low accuracy and training efficiency for large datasets due to neglecting viewwise intercorrelation, not considering diverse correlations, and high computation burden. To address these issues, a novel framework called fast broad M3L (FBM3L) is proposed, which utilizes viewwise intercorrelation, achieves joint learning among diverse correlations, and significantly reduces training time. Experiments show that FBM3L is highly competitive in evaluation metrics (up to 64% in average precision) and much faster than most methods (up to 1030 times) on large multiview datasets.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Jie Du, Xiaoci Zhang, Peng Liu, Chi-Man Vong, Tianfu Wang
Summary: Deep metric learning (DML) has been widely used in various tasks for extracting discriminant features. However, two class-imbalance learning (CIL) problems, data scarcity and data density, can lead to misclassification. Existing DML and CIL losses fail to address these issues effectively. To mitigate these challenges, we propose an IDID loss that generates diverse features within classes and preserves the semantic correlations between classes.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Qi Lai, Jianhang Zhou, Yanfen Gan, Chi-Man Vong, C. L. Philip Chen
Summary: Multi-instance multi-label learning (MIML) problems have been extensively studied in real applications, but existing methods suffer from low accuracy and training efficiency. To address these issues, a new single-stage framework called broad multi-instance multi-label learning (BMIML) is proposed, which can simultaneously learn diverse inter-correlations between whole images, instances, and labels in single stage for higher classification accuracy and faster training time through innovative modules such as auto-weighted label enhancement learning, scalable multi-instance probabilistic regression, and interactive decision optimization.
IEEE TRANSACTIONS ON EMERGING TOPICS IN COMPUTATIONAL INTELLIGENCE
(2023)
Article
Computer Science, Artificial Intelligence
Jie Du, Peng Liu, Chi-Man Vong, Chuangquan Chen, Tianfu Wang, C. L. Philip Chen
Summary: Machine learning aims to generate predictive models from training datasets, but many real-world applications involve continuous arrival of new data, making class-incremental learning (CIL) necessary. Most current CIL methods are based on computationally expensive deep models and have issues with forgetting old knowledge. This article proposes a fast and efficient broad learning system-based CIL (BLS-CIL) method that retains old class knowledge and achieves high accuracy.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Jintao Huang, Chi-Man Vong, C. L. Philip Chen, Yimin Zhou
Summary: This paper proposes a novel multi-label classifier based on a broad learning system (BLS-MLL). It improves the classification performance and training efficiency of large-scale multi-label learning by introducing kernel-based feature reduction and correlation-based label thresholding.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)