Pub Date : 2024-08-20DOI: 10.1109/TETCI.2024.3406691
Linhao Zhang;Li Jin;Guangluan Xu;Xiaoyu Li;Xian Sun
Counter-narratives, which are direct responses consisting of non-aggressive fact-based arguments, have emerged as a highly effective approach to combat the proliferation of hate speech. Previous methodologies have primarily focused on fine-tuning and post-editing techniques to ensure the fluency of generated contents, while overlooking the critical aspects of individualization and relevance concerning the specific hatred targets, such as LGBT groups, immigrants, etc. This research paper introduces a novel framework based on contrastive optimal transport, which effectively addresses the challenges of maintaining target interaction and promoting diversification in generating counter-narratives. Firstly, an Optimal Transport Kernel (OTK) module is leveraged to incorporate hatred target information in the token representations, in which the comparison pairs are extracted between original and transported features. Secondly, a self-contrastive learning module is employed to address the issue of model degeneration. This module achieves this by generating an anisotropic distribution of token representations. Finally, a target-oriented search method is integrated as an improved decoding strategy to explicitly promote domain relevance and diversification in the inference process. This strategy modifies the model's confidence score by considering both token similarity and target relevance. Quantitative and qualitative experiments have been evaluated on two benchmark datasets, which demonstrate that our proposed model significantly outperforms current methods evaluated by metrics from multiple aspects.
{"title":"COT: A Generative Approach for Hate Speech Counter-Narratives via Contrastive Optimal Transport","authors":"Linhao Zhang;Li Jin;Guangluan Xu;Xiaoyu Li;Xian Sun","doi":"10.1109/TETCI.2024.3406691","DOIUrl":"https://doi.org/10.1109/TETCI.2024.3406691","url":null,"abstract":"Counter-narratives, which are direct responses consisting of non-aggressive fact-based arguments, have emerged as a highly effective approach to combat the proliferation of hate speech. Previous methodologies have primarily focused on fine-tuning and post-editing techniques to ensure the fluency of generated contents, while overlooking the critical aspects of individualization and relevance concerning the specific hatred targets, such as LGBT groups, immigrants, etc. This research paper introduces a novel framework based on contrastive optimal transport, which effectively addresses the challenges of maintaining target interaction and promoting diversification in generating counter-narratives. Firstly, an Optimal Transport Kernel (OTK) module is leveraged to incorporate hatred target information in the token representations, in which the comparison pairs are extracted between original and transported features. Secondly, a self-contrastive learning module is employed to address the issue of model degeneration. This module achieves this by generating an anisotropic distribution of token representations. Finally, a target-oriented search method is integrated as an improved decoding strategy to explicitly promote domain relevance and diversification in the inference process. This strategy modifies the model's confidence score by considering both token similarity and target relevance. Quantitative and qualitative experiments have been evaluated on two benchmark datasets, which demonstrate that our proposed model significantly outperforms current methods evaluated by metrics from multiple aspects.","PeriodicalId":13135,"journal":{"name":"IEEE Transactions on Emerging Topics in Computational Intelligence","volume":"9 1","pages":"740-756"},"PeriodicalIF":5.3,"publicationDate":"2024-08-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143106868","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Advanced language models demonstrate remarkable capabilities but remain vulnerable to adversarial word camouflage techniques. These techniques introduce visually perceptible language manipulations while conveying intended meanings to the target audience, potentially altering a model's output. This study explores the effectiveness and limitations of word camouflage in deceiving various language model architectures, including encoder-decoder, encoder-only, and decoder-only models, with a significant focus on the tokenizers employed. Despite their vocabulary diversity, all tokenizers exhibited notable weaknesses against camouflaged words, particularly keyword attacks, highlighting the urgent need for adaptable tokenizers to handle sophisticated adversarial strategies. Consistent with our findings, transformer models without specific training against word camouflage become increasingly compromised as the complexity and volume of camouflaged inputs grow. To address these vulnerabilities, we evaluated external countermeasures such as MASK and BLANK filters, demonstrating that semantic content persists in camouflaged text and can be exploited by models. We employed static and dynamic adversarial training methods, with static training introducing camouflaged data once, while dynamic training continuously updates the data during training. Our results showed that dynamic training effectively counters adversarial attacks and enhances overall model performance, suggesting its dual role as a defensive mechanism and a data augmentation technique. The methodology incorporates the AugLy library for external validation, demonstrating the superior efficacy of dynamic training over static methods. Key contributions include enhancing the open-source tool pyleetspeak to facilitate the creation of augmented camouflaged datasets, providing researchers and practitioners with effective tools to strengthen NLP systems against evolving threats in digital communication.
{"title":"Camouflage Is All You Need: Evaluating and Enhancing Transformer Models Robustness Against Camouflage Adversarial Attacks","authors":"Álvaro Huertas-García;Alejandro Martín;Javier Huertas-Tato;David Camacho","doi":"10.1109/TETCI.2024.3440181","DOIUrl":"https://doi.org/10.1109/TETCI.2024.3440181","url":null,"abstract":"Advanced language models demonstrate remarkable capabilities but remain vulnerable to adversarial word camouflage techniques. These techniques introduce visually perceptible language manipulations while conveying intended meanings to the target audience, potentially altering a model's output. This study explores the effectiveness and limitations of word camouflage in deceiving various language model architectures, including encoder-decoder, encoder-only, and decoder-only models, with a significant focus on the tokenizers employed. Despite their vocabulary diversity, all tokenizers exhibited notable weaknesses against camouflaged words, particularly keyword attacks, highlighting the urgent need for adaptable tokenizers to handle sophisticated adversarial strategies. Consistent with our findings, transformer models without specific training against word camouflage become increasingly compromised as the complexity and volume of camouflaged inputs grow. To address these vulnerabilities, we evaluated external countermeasures such as MASK and BLANK filters, demonstrating that semantic content persists in camouflaged text and can be exploited by models. We employed static and dynamic adversarial training methods, with static training introducing camouflaged data once, while dynamic training continuously updates the data during training. Our results showed that dynamic training effectively counters adversarial attacks and enhances overall model performance, suggesting its dual role as a defensive mechanism and a data augmentation technique. The methodology incorporates the AugLy library for external validation, demonstrating the superior efficacy of dynamic training over static methods. Key contributions include enhancing the open-source tool pyleetspeak to facilitate the creation of augmented camouflaged datasets, providing researchers and practitioners with effective tools to strengthen NLP systems against evolving threats in digital communication.","PeriodicalId":13135,"journal":{"name":"IEEE Transactions on Emerging Topics in Computational Intelligence","volume":"9 1","pages":"431-443"},"PeriodicalIF":5.3,"publicationDate":"2024-08-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143361538","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-07-29DOI: 10.1109/TETCI.2024.3423472
Ting-En Chao;Yu Huang;Hao Dai;Gary G. Yen;Vincent S. Tseng
Anomaly prediction, aiming to predict abnormal events before occurrence, plays a key role in significantly reducing costs and minimizing potential threats to mechanical devices. Monitoring machines using fixed-length time windows faces challenges in accommodating the varying characteristics of anomaly events. The lengthy and imbalanced sequence data associated with anomaly events further complicates the resolution of these challenges. Additionally, the inherent trade-off between accurate prediction and timely alarm is a crucial concern, posing difficulties in decision-making. This study puts forward a novel framework for early anomaly prediction, named early time series Anomaly Prediction with Neighbor Over-sampling and Multi-objective Optimization (APNOMO), pronounced as ‘abnormal’. The framework employs three key techniques: 1) sliding windows that divide long input sequences into segments for prediction at proper intervals, 2) over-sampling and proposed neighbor over-sampling that handle imbalanced data in a novel way, and 3) multi-objective optimization that searches optimal thresholds to balance accurately prediction and timely alarm the abnormal. Experiments on a real-world dataset demonstrate APNOMO's superior performance over some state-of-the-art designs, with higher recall, F1 score, and more suitable earliness. It can predict anomalies 0.78–3.56 hours in advance, showcasing excellent early anomaly prediction capabilities for enabling predictive maintenance.
{"title":"Early Time Series Anomaly Prediction With Multi-Objective Optimization","authors":"Ting-En Chao;Yu Huang;Hao Dai;Gary G. Yen;Vincent S. Tseng","doi":"10.1109/TETCI.2024.3423472","DOIUrl":"https://doi.org/10.1109/TETCI.2024.3423472","url":null,"abstract":"Anomaly prediction, aiming to predict abnormal events before occurrence, plays a key role in significantly reducing costs and minimizing potential threats to mechanical devices. Monitoring machines using fixed-length time windows faces challenges in accommodating the varying characteristics of anomaly events. The lengthy and imbalanced sequence data associated with anomaly events further complicates the resolution of these challenges. Additionally, the inherent trade-off between accurate prediction and timely alarm is a crucial concern, posing difficulties in decision-making. This study puts forward a novel framework for early anomaly prediction, named early time series Anomaly Prediction with Neighbor Over-sampling and Multi-objective Optimization (APNOMO), pronounced as ‘abnormal’. The framework employs three key techniques: 1) sliding windows that divide long input sequences into segments for prediction at proper intervals, 2) over-sampling and proposed neighbor over-sampling that handle imbalanced data in a novel way, and 3) multi-objective optimization that searches optimal thresholds to balance accurately prediction and timely alarm the abnormal. Experiments on a real-world dataset demonstrate APNOMO's superior performance over some state-of-the-art designs, with higher recall, F1 score, and more suitable earliness. It can predict anomalies 0.78–3.56 hours in advance, showcasing excellent early anomaly prediction capabilities for enabling predictive maintenance.","PeriodicalId":13135,"journal":{"name":"IEEE Transactions on Emerging Topics in Computational Intelligence","volume":"9 1","pages":"972-987"},"PeriodicalIF":5.3,"publicationDate":"2024-07-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143106796","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Chest X-ray images play a crucial role in pneumonia diagnosis, with deep transfer learning being a widely adopted method for pneumonia detection. However, effectively handling feature data extracted from deep models without succumbing to the challenges of feature dimensionality remains a formidable task. In response to this complex issue, we propose a novel two-stage deep feature selection (FS) method utilizing the voting differential evolution (VDE) algorithm. In this approach, a dimension adaptive search strategy is meticulously devised to ensure robust feature selection while concurrently reducing the dimension. To expedite the optimization process, we devise a CR adaptive adjustment method to enhance the efficiency of the algorithm. Notably, an important aspect of our approach is the introduction of a novel DE algorithm that integrates a voting mechanism. This synergistic fusion allows a comprehensive analysis of crucial feature relationships to mitigate the risk of algorithmic entrapment in local optima. Additionally, we propose a dynamic feature evaluation function to avert the oversight of feature sets with optimal classification accuracy during later stages of the algorithm, thereby preserving discriminative features. The method is verified on an open Chest X-Ray Images dataset, achieving 99.04% average precision, 98.67% average accuracy, 99.13% average recall, and 19.93% average feature dimension reduction ratio. The experimental findings reveal that the presented method outperform prevailing state-of-the-art algorithms.
{"title":"Two-Stage Deep Feature Selection Method Using Voting Differential Evolution Algorithm for Pneumonia Detection From Chest X-Ray Images","authors":"Haibin Ouyang;Dongmei Liu;Steven Li;Weiping Ding;Zhi-Hui Zhan","doi":"10.1109/TETCI.2024.3425285","DOIUrl":"https://doi.org/10.1109/TETCI.2024.3425285","url":null,"abstract":"Chest X-ray images play a crucial role in pneumonia diagnosis, with deep transfer learning being a widely adopted method for pneumonia detection. However, effectively handling feature data extracted from deep models without succumbing to the challenges of feature dimensionality remains a formidable task. In response to this complex issue, we propose a novel two-stage deep feature selection (FS) method utilizing the voting differential evolution (VDE) algorithm. In this approach, a dimension adaptive search strategy is meticulously devised to ensure robust feature selection while concurrently reducing the dimension. To expedite the optimization process, we devise a CR adaptive adjustment method to enhance the efficiency of the algorithm. Notably, an important aspect of our approach is the introduction of a novel DE algorithm that integrates a voting mechanism. This synergistic fusion allows a comprehensive analysis of crucial feature relationships to mitigate the risk of algorithmic entrapment in local optima. Additionally, we propose a dynamic feature evaluation function to avert the oversight of feature sets with optimal classification accuracy during later stages of the algorithm, thereby preserving discriminative features. The method is verified on an open Chest X-Ray Images dataset, achieving 99.04% average precision, 98.67% average accuracy, 99.13% average recall, and 19.93% average feature dimension reduction ratio. The experimental findings reveal that the presented method outperform prevailing state-of-the-art algorithms.","PeriodicalId":13135,"journal":{"name":"IEEE Transactions on Emerging Topics in Computational Intelligence","volume":"9 1","pages":"918-932"},"PeriodicalIF":5.3,"publicationDate":"2024-07-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143106864","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-07-23DOI: 10.1109/TETCI.2024.3427473
{"title":"IEEE Computational Intelligence Society Information","authors":"","doi":"10.1109/TETCI.2024.3427473","DOIUrl":"https://doi.org/10.1109/TETCI.2024.3427473","url":null,"abstract":"","PeriodicalId":13135,"journal":{"name":"IEEE Transactions on Emerging Topics in Computational Intelligence","volume":"8 4","pages":"C3-C3"},"PeriodicalIF":5.3,"publicationDate":"2024-07-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10607836","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141964719","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Utilizing data acquired by multiple wearable sensors can usually guarantee more accurate recognition for deep learning based human activity recognition. However, an increased number of sensors bring high processing cost, influencing real-time activity monitoring. Besides, existing methods rarely consider the interpretability of the recognition model in aspects of both the importance of the sensors and features, causing a gap between deep learning and their extendability in real-world scenario. In this paper, we cast the classical fused lasso model into a deep neural network, proposing a deep fused Lasso net (dfLasso-Net), which can perform sensor selection, feature selection and HAR in one end-to-end structure. Specifically, a two-level weight computing module (TLWCM) consisting of a senor weight net and a feature weight net is designed to measure the importance of sensors and features. In sensor weight net, spatial smoothness between physical channels within each sensor is considered to maximize the usage of selected sensors. And the feature weight net is able to maintain the physical meaning of the hand-crafted features through feature selection inside the sensors. By combining with the learning module for classification, HAR can be performed. We test dfLasso-Net on three multi-sensor based HAR datasets, demonstrating that dfLasso-Net achieves better recognition accuracy with the least number of sensors and provides good model interpretability by visualizing the weights of the sensors and features. Last but not least, dflasso-Net can be used as an effective filter-based feature selection approach with much flexibility.
利用多个可穿戴传感器获取的数据通常可以保证基于深度学习的人类活动识别更加准确。然而，传感器数量的增加会带来高昂的处理成本，影响实时活动监测。此外，现有方法很少从传感器和特征的重要性两方面考虑识别模型的可解释性，导致深度学习与其在实际场景中的可扩展性之间存在差距。本文将经典的融合拉索模型转化为深度神经网络，提出了一种深度融合拉索网络（dfLasso-Net），它可以在一个端到端的结构中完成传感器选择、特征选择和HAR。具体来说，设计了一个由传感器权重网和特征权重网组成的两级权重计算模块（TLWCM）来衡量传感器和特征的重要性。在传感器权重网中，考虑了每个传感器内物理通道之间的空间平滑性，以最大限度地提高所选传感器的使用率。而特征权重网能够通过传感器内部的特征选择，保持手工创建特征的物理意义。通过与用于分类的学习模块相结合，可以执行 HAR。我们在三个基于多传感器的 HAR 数据集上测试了 dfLasso-Net，结果表明 dfLassoNet 以最少的传感器数量实现了更高的识别准确率，并通过可视化传感器和特征的权重提供了良好的模型可解释性。最后但并非最不重要的一点是，dflasso-Net 可以作为一种有效的基于滤波器的特征选择方法，具有很大的灵活性。
{"title":"Energy-Efficient and Interpretable Multisensor Human Activity Recognition via Deep Fused Lasso Net","authors":"Yu Zhou;Jingtao Xie;Xiao Zhang;Wenhui Wu;Sam Kwong","doi":"10.1109/TETCI.2024.3430008","DOIUrl":"https://doi.org/10.1109/TETCI.2024.3430008","url":null,"abstract":"Utilizing data acquired by multiple wearable sensors can usually guarantee more accurate recognition for deep learning based human activity recognition. However, an increased number of sensors bring high processing cost, influencing real-time activity monitoring. Besides, existing methods rarely consider the interpretability of the recognition model in aspects of both the importance of the sensors and features, causing a gap between deep learning and their extendability in real-world scenario. In this paper, we cast the classical fused lasso model into a deep neural network, proposing a deep fused Lasso net (dfLasso-Net), which can perform sensor selection, feature selection and HAR in one end-to-end structure. Specifically, a two-level weight computing module (TLWCM) consisting of a senor weight net and a feature weight net is designed to measure the importance of sensors and features. In sensor weight net, spatial smoothness between physical channels within each sensor is considered to maximize the usage of selected sensors. And the feature weight net is able to maintain the physical meaning of the hand-crafted features through feature selection inside the sensors. By combining with the learning module for classification, HAR can be performed. We test dfLasso-Net on three multi-sensor based HAR datasets, demonstrating that dfLasso-Net achieves better recognition accuracy with the least number of sensors and provides good model interpretability by visualizing the weights of the sensors and features. Last but not least, dflasso-Net can be used as an effective filter-based feature selection approach with much flexibility.","PeriodicalId":13135,"journal":{"name":"IEEE Transactions on Emerging Topics in Computational Intelligence","volume":"8 5","pages":"3576-3588"},"PeriodicalIF":5.3,"publicationDate":"2024-07-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142376908","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-07-23DOI: 10.1109/TETCI.2024.3427471
{"title":"IEEE Transactions on Emerging Topics in Computational Intelligence Publication Information","authors":"","doi":"10.1109/TETCI.2024.3427471","DOIUrl":"https://doi.org/10.1109/TETCI.2024.3427471","url":null,"abstract":"","PeriodicalId":13135,"journal":{"name":"IEEE Transactions on Emerging Topics in Computational Intelligence","volume":"8 4","pages":"C2-C2"},"PeriodicalIF":5.3,"publicationDate":"2024-07-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10607837","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141965822","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-07-23DOI: 10.1109/TETCI.2024.3427475
{"title":"IEEE Transactions on Emerging Topics in Computational Intelligence Information for Authors","authors":"","doi":"10.1109/TETCI.2024.3427475","DOIUrl":"https://doi.org/10.1109/TETCI.2024.3427475","url":null,"abstract":"","PeriodicalId":13135,"journal":{"name":"IEEE Transactions on Emerging Topics in Computational Intelligence","volume":"8 4","pages":"C4-C4"},"PeriodicalIF":5.3,"publicationDate":"2024-07-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10607838","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141965032","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-07-22DOI: 10.1109/TETCI.2024.3420735
Yurong Zhong;Kechen Liu;Shangce Gao;Xin Luo
Large scale interaction data are frequently found in industrial applications related with Big Data. Due to the fact that few interactions commonly happen among numerous nodes in real scenes, such data can be quantified into a High-Dimensional and Incomplete (HDI) matrix where most entries are unknown. An alternating-direction-method-based nonnegative latent factor model can perform efficient and accurate representation leaning to an HDI matrix, while its multiple hyper-parameters greatly limit its scalability for real applications. Aiming at implementing a highly-scalable and efficient latent factor model, this paper adopts the principle of particle swarm optimization and the tree-structured parzen estimator algorithm to facilitate the hyper-parameter adaptation mechanism, thereby building an Alternating-direction-method-based Adaptive Nonnegative Latent Factor (A 2 NLF) model. Its theoretical convergence is rigorously proved. Empirical studies on several nonnegative HDI matrices from real applications demonstrate that the proposed A 2 NLF model obtains higher computational and storage efficiency than several state-of-the-art models, along with significant accuracy gain. Its hyper-parameter adaptation is implemented smoothly, thereby greatly boosting its scalability in real problems.
{"title":"Alternating-Direction-Method of Multipliers-Based Adaptive Nonnegative Latent Factor Analysis","authors":"Yurong Zhong;Kechen Liu;Shangce Gao;Xin Luo","doi":"10.1109/TETCI.2024.3420735","DOIUrl":"https://doi.org/10.1109/TETCI.2024.3420735","url":null,"abstract":"Large scale interaction data are frequently found in industrial applications related with Big Data. Due to the fact that few interactions commonly happen among numerous nodes in real scenes, such data can be quantified into a High-Dimensional and Incomplete (HDI) matrix where most entries are unknown. An alternating-direction-method-based nonnegative latent factor model can perform efficient and accurate representation leaning to an HDI matrix, while its multiple hyper-parameters greatly limit its scalability for real applications. Aiming at implementing a highly-scalable and efficient latent factor model, this paper adopts the principle of particle swarm optimization and the tree-structured parzen estimator algorithm to facilitate the hyper-parameter adaptation mechanism, thereby building an Alternating-direction-method-based Adaptive Nonnegative Latent Factor (A\u0000<sup>2</sup>\u0000NLF) model. Its theoretical convergence is rigorously proved. Empirical studies on several nonnegative HDI matrices from real applications demonstrate that the proposed A\u0000<sup>2</sup>\u0000NLF model obtains higher computational and storage efficiency than several state-of-the-art models, along with significant accuracy gain. Its hyper-parameter adaptation is implemented smoothly, thereby greatly boosting its scalability in real problems.","PeriodicalId":13135,"journal":{"name":"IEEE Transactions on Emerging Topics in Computational Intelligence","volume":"8 5","pages":"3544-3558"},"PeriodicalIF":5.3,"publicationDate":"2024-07-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142376995","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
In complex environments where occupied and unknown areas exceed the free space, it is essential for robots to utilize efficient methods for environmental perception, trajectory planning, and trajectory tracking. This paper introduces the jump point search (JPS) algorithm as a global planning approach and integrates the complete trajectory and safe trajectory using convex decomposition for local planning purposes. We specifically formulate the planning process as a jerk optimization problem to reduce robot vibrations and improve stability. To address trajectory tracking challenges, we propose an innovative robust control Lyapunov function method. This method efficiently manages disturbances in mobile robot motion, enhancing stability. It considers input constraints such as angular and linear velocity limits, along with optimization metrics like minimal input effort. We utilize a proximal augmented Lagrangian method to solve the optimization problem related to trajectory planning and the robust control Lyapunov function. Through experiments involving different friction forces and torques, we validate the effectiveness of our proposed robust control Lyapunov function controller in managing unknown disturbances. This demonstrates its superior adaptability and robustness compared to conventional model control.
{"title":"Efficient Online Planning and Robust Optimal Control for Nonholonomic Mobile Robot in Unstructured Environments","authors":"Yingbai Hu;Wei Zhou;Yueyue Liu;Minghao Zeng;Weiping Ding;Shu Li;Guoxin Li;Zheng Li;Alois Knoll","doi":"10.1109/TETCI.2024.3424527","DOIUrl":"https://doi.org/10.1109/TETCI.2024.3424527","url":null,"abstract":"In complex environments where occupied and unknown areas exceed the free space, it is essential for robots to utilize efficient methods for environmental perception, trajectory planning, and trajectory tracking. This paper introduces the jump point search (JPS) algorithm as a global planning approach and integrates the complete trajectory and safe trajectory using convex decomposition for local planning purposes. We specifically formulate the planning process as a jerk optimization problem to reduce robot vibrations and improve stability. To address trajectory tracking challenges, we propose an innovative robust control Lyapunov function method. This method efficiently manages disturbances in mobile robot motion, enhancing stability. It considers input constraints such as angular and linear velocity limits, along with optimization metrics like minimal input effort. We utilize a proximal augmented Lagrangian method to solve the optimization problem related to trajectory planning and the robust control Lyapunov function. Through experiments involving different friction forces and torques, we validate the effectiveness of our proposed robust control Lyapunov function controller in managing unknown disturbances. This demonstrates its superior adaptability and robustness compared to conventional model control.","PeriodicalId":13135,"journal":{"name":"IEEE Transactions on Emerging Topics in Computational Intelligence","volume":"8 5","pages":"3559-3575"},"PeriodicalIF":5.3,"publicationDate":"2024-07-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142377138","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：