In the post-genomic era, a central challenge for disease genomes is the identification of the biological effects of specific somatic variants on allosteric proteins and the phenotypes they influence during the initiation and progression of diseases. Here, we analyze more than 38 539 mutations observed in 90 human genes with 740 allosteric protein chains. We find that existing allosteric protein mutations are associated with many diseases, but the clinical significance of most mutations in allosteric proteins remains unclear. Next, we develop an ensemble-learning-based model for pathogenic mutation prediction of allosteric proteins based on the intrinsic characteristics of proteins and the prediction results from existed methods. When tested on the benchmark allosteric protein dataset, the proposed method achieves an AUCs of 0.868 and an AUPR of 0.894 on allosteric proteins. Furthermore, we explore the performance of existing methods in predicting the pathogenicity of mutations at allosteric sites and identify potential significant pathogenic mutations at allosteric sites using the proposed method. In summary, these findings illuminate the significance of allosteric mutation in disease processes, and contribute a valuable tool for the identification of pathogenic mutations as well as previously unknown disease-causing allosteric-protein-encoded genes.
{"title":"Exploring Pathogenic Mutation in Allosteric Proteins: The Prediction and Beyond","authors":"Huiling Zhang;Zhen Ju;Jingjing Zhang;Xijian Li;Hanyang Xiao;Xiaochuan Chen;Yuetong Li;Xinran Wang;Yanjie Wei","doi":"10.26599/TST.2024.9010226","DOIUrl":"https://doi.org/10.26599/TST.2024.9010226","url":null,"abstract":"In the post-genomic era, a central challenge for disease genomes is the identification of the biological effects of specific somatic variants on allosteric proteins and the phenotypes they influence during the initiation and progression of diseases. Here, we analyze more than 38 539 mutations observed in 90 human genes with 740 allosteric protein chains. We find that existing allosteric protein mutations are associated with many diseases, but the clinical significance of most mutations in allosteric proteins remains unclear. Next, we develop an ensemble-learning-based model for pathogenic mutation prediction of allosteric proteins based on the intrinsic characteristics of proteins and the prediction results from existed methods. When tested on the benchmark allosteric protein dataset, the proposed method achieves an AUCs of 0.868 and an AUPR of 0.894 on allosteric proteins. Furthermore, we explore the performance of existing methods in predicting the pathogenicity of mutations at allosteric sites and identify potential significant pathogenic mutations at allosteric sites using the proposed method. In summary, these findings illuminate the significance of allosteric mutation in disease processes, and contribute a valuable tool for the identification of pathogenic mutations as well as previously unknown disease-causing allosteric-protein-encoded genes.","PeriodicalId":48690,"journal":{"name":"Tsinghua Science and Technology","volume":"30 5","pages":"2284-2299"},"PeriodicalIF":6.6,"publicationDate":"2025-04-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10979792","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143888380","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"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}
The computation of matrix pseudoinverses is a recurrent requirement across various scientific computing and engineering domains. The prevailing models for matrix pseudoinverse typically operate under the assumption of a noise-free solution process or presume that any noise present has been effectively addressed prior to computation. However, the concurrent real-time computation of time-varying matrix pseudoinverses holds significant practical utility, while the preemptive preprocessing for noise elimination or reduction may impose supplementary computational overheads on real-time implementations. Different from previous models for solving the pseudoinverse of time-varying matrices, in this paper, a model for solving the pseudoinverse of time-varying matrices using a double-integral structure, called Double-Integral-Enhanced Zeroing Neural Network (DIEZNN) model, is proposed and investigated, which is capable of solving time-varying matrix pseudoinverse while efficiently eliminating the negative effects of linear noise perturbations. The experimental results show that in the presence of linear noise, the DIEZNN model demonstrates better noise suppression performance compared to both the original zeroing neural network model and the Zeroing Neural Network (ZNN) model enhanced with a Li-type activation function. In addition, these models are applied to the control of chaotic system of controllable permanent magnet synchronous motor, which further verifies the superiority of DIEZNN in engineering application.
{"title":"A Novel Zeroing Neural Network for Time-Varying Matrix Pseudoinversion in the Presence of Linear Noises","authors":"Jianfeng Li;Linxi Qu;Yueming Zhu;Zhan Li;Bolin Liao","doi":"10.26599/TST.2024.9010120","DOIUrl":"https://doi.org/10.26599/TST.2024.9010120","url":null,"abstract":"The computation of matrix pseudoinverses is a recurrent requirement across various scientific computing and engineering domains. The prevailing models for matrix pseudoinverse typically operate under the assumption of a noise-free solution process or presume that any noise present has been effectively addressed prior to computation. However, the concurrent real-time computation of time-varying matrix pseudoinverses holds significant practical utility, while the preemptive preprocessing for noise elimination or reduction may impose supplementary computational overheads on real-time implementations. Different from previous models for solving the pseudoinverse of time-varying matrices, in this paper, a model for solving the pseudoinverse of time-varying matrices using a double-integral structure, called Double-Integral-Enhanced Zeroing Neural Network (DIEZNN) model, is proposed and investigated, which is capable of solving time-varying matrix pseudoinverse while efficiently eliminating the negative effects of linear noise perturbations. The experimental results show that in the presence of linear noise, the DIEZNN model demonstrates better noise suppression performance compared to both the original zeroing neural network model and the Zeroing Neural Network (ZNN) model enhanced with a Li-type activation function. In addition, these models are applied to the control of chaotic system of controllable permanent magnet synchronous motor, which further verifies the superiority of DIEZNN in engineering application.","PeriodicalId":48690,"journal":{"name":"Tsinghua Science and Technology","volume":"30 5","pages":"1911-1926"},"PeriodicalIF":6.6,"publicationDate":"2025-04-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10979798","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143888379","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"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}
Most traffic accidents are caused by improper driver operation, so autonomous driving based on rapidly developing artificial intelligence technology has attracted much attention. Inspired by the biological visual perception and neural decision-making mechanism, this paper constructs a two-branch bionic network for autonomous driving, which learns to map the driver's perspective image directly to the steering commands. On the real-world driving dataset we collected, extensive experiments prove the efficiency, robustness, superior structure and biological interpretability of this end-to-end algorithm. Moreover, the flexible scalability of this network greatly supports real-time inference and deployment.
{"title":"End-to-End Two-Branch Bionic Network for Autonomous Driving","authors":"Guoliang Sun;Sifa Zheng;Xingrui Gong;Yijie Pan;Rui Yang;Yingying Yu;Shanshan Pei","doi":"10.26599/TST.2024.9010170","DOIUrl":"https://doi.org/10.26599/TST.2024.9010170","url":null,"abstract":"Most traffic accidents are caused by improper driver operation, so autonomous driving based on rapidly developing artificial intelligence technology has attracted much attention. Inspired by the biological visual perception and neural decision-making mechanism, this paper constructs a two-branch bionic network for autonomous driving, which learns to map the driver's perspective image directly to the steering commands. On the real-world driving dataset we collected, extensive experiments prove the efficiency, robustness, superior structure and biological interpretability of this end-to-end algorithm. Moreover, the flexible scalability of this network greatly supports real-time inference and deployment.","PeriodicalId":48690,"journal":{"name":"Tsinghua Science and Technology","volume":"30 5","pages":"2259-2269"},"PeriodicalIF":6.6,"publicationDate":"2025-04-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10979786","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143888403","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"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 : 2025-04-29DOI: 10.26599/TST.2024.9010145
Changxian Xu;Jiliang Zhang;Keping Liu;Jian Wang;Zhongbo Sun
In this paper, a Novel Compliant Actuator (NCA)-driven Upper-Limb Exoskeleton (ULE) with force controllable, impact resistance, and back drivability is designed to ensure the safety of the subject during Human-Robot Interaction (HRI) processing. Based on the designed NCA-driven ULE, this paper constructs a Model Predictive Control Scheme (MPCS) for force trajectory tracking, which minimises future tracking errors by solving an optimal control problem with inequality constraints. In addition, an Error-Accumulation Improved Newton Algorithm (EAINA) is proposed to solve the MPCS for suppressing various noises and external disturbances. The proposed EAINA is theoretically proved to have small steady state for noise conditions and stability of the EAINA using Lyapunov method. Finally, experimental results verify that the proposed MPCS solved by the EAINA in the NCA-driven ULE achieves robustness, fast convergence, strong tolerance and stability for trajectory rehabilitation task.
{"title":"Error-Accumulation Improved Newton Algorithm in Model Predictive Control for Novel Compliant Actuator-Driven Upper-Limb Exoskeleton","authors":"Changxian Xu;Jiliang Zhang;Keping Liu;Jian Wang;Zhongbo Sun","doi":"10.26599/TST.2024.9010145","DOIUrl":"https://doi.org/10.26599/TST.2024.9010145","url":null,"abstract":"In this paper, a Novel Compliant Actuator (NCA)-driven Upper-Limb Exoskeleton (ULE) with force controllable, impact resistance, and back drivability is designed to ensure the safety of the subject during Human-Robot Interaction (HRI) processing. Based on the designed NCA-driven ULE, this paper constructs a Model Predictive Control Scheme (MPCS) for force trajectory tracking, which minimises future tracking errors by solving an optimal control problem with inequality constraints. In addition, an Error-Accumulation Improved Newton Algorithm (EAINA) is proposed to solve the MPCS for suppressing various noises and external disturbances. The proposed EAINA is theoretically proved to have small steady state for noise conditions and stability of the EAINA using Lyapunov method. Finally, experimental results verify that the proposed MPCS solved by the EAINA in the NCA-driven ULE achieves robustness, fast convergence, strong tolerance and stability for trajectory rehabilitation task.","PeriodicalId":48690,"journal":{"name":"Tsinghua Science and Technology","volume":"30 5","pages":"1965-1979"},"PeriodicalIF":6.6,"publicationDate":"2025-04-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10979799","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143888409","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"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 : 2025-04-29DOI: 10.26599/TST.2024.9010117
Bingbin Liu;Yuxuan Qian;Jianxin Wang
Although multi-view monitoring techniques have been widely applied in skinned model reconstruction and movement analysis, traditional systems using high-performance Personal Computers (PCs), or industrial cameras are often prohibitive due to high costs and limited scalability. Here, we introduce an affordable, scalable multi-view image acquisition system for skinned model reconstruction in animal studies, utilizing consumer Android devices and a wireless network for synchronized monitoring named Rodent Arena Multi-View Monitor (RAMM). It uses smartphones as camera nodes with local data storage, enabling cost-effective scalability. Its custom synchronization solution and portability make it ideal for research and education in rodent behavior analysis, offering a practical alternative for institutions with limited budgets. Furthermore, the portability and flexibility of this system make it an ideal tool for rodent skinned model research based on multi-view image acquisition. To evaluate the performance, we perform an oscilloscope analysis to ensure effectiveness of synchronization. A 45-camera node setup is built to highlight RAMM's cost efficiency and ease in constructing large-scale systems. Additionally, the data quality is validated using the Instant Neural Graphics Primitives (Instant-NGP) method. Remarkable results were achieved with a 30.49 dB PSNR by utilizing only 25 images with intrinsic and extrinsic parameters, fulfilling the requirements for well-synchronized data used in 3D representation algorithms.
{"title":"Rodent Arena Multi-View Monitor (RAMM): A Camera Synchronized Photographic Control System for Multi-View Rodent Monitoring","authors":"Bingbin Liu;Yuxuan Qian;Jianxin Wang","doi":"10.26599/TST.2024.9010117","DOIUrl":"https://doi.org/10.26599/TST.2024.9010117","url":null,"abstract":"Although multi-view monitoring techniques have been widely applied in skinned model reconstruction and movement analysis, traditional systems using high-performance Personal Computers (PCs), or industrial cameras are often prohibitive due to high costs and limited scalability. Here, we introduce an affordable, scalable multi-view image acquisition system for skinned model reconstruction in animal studies, utilizing consumer Android devices and a wireless network for synchronized monitoring named Rodent Arena Multi-View Monitor (RAMM). It uses smartphones as camera nodes with local data storage, enabling cost-effective scalability. Its custom synchronization solution and portability make it ideal for research and education in rodent behavior analysis, offering a practical alternative for institutions with limited budgets. Furthermore, the portability and flexibility of this system make it an ideal tool for rodent skinned model research based on multi-view image acquisition. To evaluate the performance, we perform an oscilloscope analysis to ensure effectiveness of synchronization. A 45-camera node setup is built to highlight RAMM's cost efficiency and ease in constructing large-scale systems. Additionally, the data quality is validated using the Instant Neural Graphics Primitives (Instant-NGP) method. Remarkable results were achieved with a 30.49 dB PSNR by utilizing only 25 images with intrinsic and extrinsic parameters, fulfilling the requirements for well-synchronized data used in 3D representation algorithms.","PeriodicalId":48690,"journal":{"name":"Tsinghua Science and Technology","volume":"30 5","pages":"2195-2214"},"PeriodicalIF":6.6,"publicationDate":"2025-04-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10979787","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143888406","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"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 : 2025-04-29DOI: 10.26599/TST.2024.9010130
Yunshan Wei;Sixian Xiong;Wenli Shang
This paper proposed a feedback feedforward Iterative Learning Control (ILC) law for nonlinear system with iteratively variable trial lengths under a networked systems structure, where the both sensor and actuator occurs random data lost separately. The feedforward ILC part includes the calculated input signal, actual input signal, and the modified tracking error of last iteration. Some tracking signal would be lost at last iteration because of the iterative varying trial lengths. In order to offset the missing signal of last trial, the tracking error of present trial is adopted by feedback control part. It is established that the convergence relied on the feedforward control gain merely, while the rate of convergence is also expedited by the feedback control component. When the initial state expectation equals to the reference one, it is established that the tracking error expectation can be controlled to zero. With an illustrative simulation, the effectiveness of the developed algorithm can be demonstrated.
{"title":"Feedback Feedforward Iterative Learning Control for Networked Nonlinear System Under Iteratively Variable Trial Lengths and Data Dropouts","authors":"Yunshan Wei;Sixian Xiong;Wenli Shang","doi":"10.26599/TST.2024.9010130","DOIUrl":"https://doi.org/10.26599/TST.2024.9010130","url":null,"abstract":"This paper proposed a feedback feedforward Iterative Learning Control (ILC) law for nonlinear system with iteratively variable trial lengths under a networked systems structure, where the both sensor and actuator occurs random data lost separately. The feedforward ILC part includes the calculated input signal, actual input signal, and the modified tracking error of last iteration. Some tracking signal would be lost at last iteration because of the iterative varying trial lengths. In order to offset the missing signal of last trial, the tracking error of present trial is adopted by feedback control part. It is established that the convergence relied on the feedforward control gain merely, while the rate of convergence is also expedited by the feedback control component. When the initial state expectation equals to the reference one, it is established that the tracking error expectation can be controlled to zero. With an illustrative simulation, the effectiveness of the developed algorithm can be demonstrated.","PeriodicalId":48690,"journal":{"name":"Tsinghua Science and Technology","volume":"30 5","pages":"1897-1910"},"PeriodicalIF":6.6,"publicationDate":"2025-04-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10979793","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143888329","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"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 : 2025-04-29DOI: 10.26599/TST.2024.9010259
Ming Sun;Xinyu Wu;Yi Zhou;Jin-Kao Hao;Zhang-Hua Fu
In wireless artificial intelligent computing systems, the construction of backbone network, which determines the optimum network for a set of given terminal nodes like users, switches, and concentrators, can be naturally formed as the Steiner tree problem. The Steiner tree problem asks for a minimum edge-weighted tree spanning a given set of terminal vertices from a given graph. As a well-known graph problem, many algorithms have been developed for solving this computationally challenging problem in the past decades. However, existing algorithms typically encounter difficulties for solving large instances, i.e., graphs with a high number of vertices and terminals. In this paper, we present a novel partition-and-merge algorithm for effectively handle large-scale graphs. The algorithm breaks the input network into small subgraphs and then merges the subgraphs in a bottom-up manner. In the merging procedure, partial Steiner trees in the subgraphs are also created and optimized by an efficient local optimization. When the merging procedure ends, the algorithm terminates and reports the final solution for the input graph. We evaluated the algorithm on a wide range of benchmark instances, showing that the algorithm outperforms the best-known algorithms on large instances and competes favorably with them on small or middle-sized instances.
{"title":"Efficient Backbone Network Construction in Wireless Artificial Intelligent Computing Systems","authors":"Ming Sun;Xinyu Wu;Yi Zhou;Jin-Kao Hao;Zhang-Hua Fu","doi":"10.26599/TST.2024.9010259","DOIUrl":"https://doi.org/10.26599/TST.2024.9010259","url":null,"abstract":"In wireless artificial intelligent computing systems, the construction of backbone network, which determines the optimum network for a set of given terminal nodes like users, switches, and concentrators, can be naturally formed as the Steiner tree problem. The Steiner tree problem asks for a minimum edge-weighted tree spanning a given set of terminal vertices from a given graph. As a well-known graph problem, many algorithms have been developed for solving this computationally challenging problem in the past decades. However, existing algorithms typically encounter difficulties for solving large instances, i.e., graphs with a high number of vertices and terminals. In this paper, we present a novel partition-and-merge algorithm for effectively handle large-scale graphs. The algorithm breaks the input network into small subgraphs and then merges the subgraphs in a bottom-up manner. In the merging procedure, partial Steiner trees in the subgraphs are also created and optimized by an efficient local optimization. When the merging procedure ends, the algorithm terminates and reports the final solution for the input graph. We evaluated the algorithm on a wide range of benchmark instances, showing that the algorithm outperforms the best-known algorithms on large instances and competes favorably with them on small or middle-sized instances.","PeriodicalId":48690,"journal":{"name":"Tsinghua Science and Technology","volume":"30 5","pages":"2300-2319"},"PeriodicalIF":6.6,"publicationDate":"2025-04-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10979815","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143888333","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"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 : 2025-04-29DOI: 10.26599/TST2024.9010134
Xin Wang;Jianhui Lv;Madini O. Alassafi;Fawaz E. Alsaadi;B. D. Parameshachari;Longhao Zou;Gang Feng;Zhonghua Liu
With the advent of deep learning, various deep neural network architectures have been proposed to capture the complex spatio-temporal dependencies in traffic data. This paper introduces a novel Deep Bi-directional Adaptive Gating Graph Convolutional Network (DBAG-GCN) model for spatio-temporal traffic forecasting. The proposed model leverages the power of graph convolutional networks to capture the spatial dependencies in the road network topology and incorporates bi-directional gating mechanisms to control the information flow adaptively. Furthermore, we introduce a multi-scale temporal convolution module to capture multi-scale temporal dynamics and a contextual attention mechanism to integrate external factors such as weather conditions and event information. Extensive experiments on real-world traffic datasets demonstrate the superior performance of DBAG-GCN compared to state-of-the-art baselines, achieving significant improvements in prediction accuracy and computational efficiency. The DBAG-GCN model provides a powerful and flexible framework for spatio-temporal traffic forecasting, paving the way for intelligent transportation management and urban planning.
{"title":"Deep Bi-Directional Adaptive Gating Graph Convolutional Networks for Spatio-Temporal Traffic Forecasting","authors":"Xin Wang;Jianhui Lv;Madini O. Alassafi;Fawaz E. Alsaadi;B. D. Parameshachari;Longhao Zou;Gang Feng;Zhonghua Liu","doi":"10.26599/TST2024.9010134","DOIUrl":"https://doi.org/10.26599/TST2024.9010134","url":null,"abstract":"With the advent of deep learning, various deep neural network architectures have been proposed to capture the complex spatio-temporal dependencies in traffic data. This paper introduces a novel Deep Bi-directional Adaptive Gating Graph Convolutional Network (DBAG-GCN) model for spatio-temporal traffic forecasting. The proposed model leverages the power of graph convolutional networks to capture the spatial dependencies in the road network topology and incorporates bi-directional gating mechanisms to control the information flow adaptively. Furthermore, we introduce a multi-scale temporal convolution module to capture multi-scale temporal dynamics and a contextual attention mechanism to integrate external factors such as weather conditions and event information. Extensive experiments on real-world traffic datasets demonstrate the superior performance of DBAG-GCN compared to state-of-the-art baselines, achieving significant improvements in prediction accuracy and computational efficiency. The DBAG-GCN model provides a powerful and flexible framework for spatio-temporal traffic forecasting, paving the way for intelligent transportation management and urban planning.","PeriodicalId":48690,"journal":{"name":"Tsinghua Science and Technology","volume":"30 5","pages":"2060-2080"},"PeriodicalIF":6.6,"publicationDate":"2025-04-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10979652","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143888435","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"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}
As mobile devices and sensor technology advance, their role in communication becomes increasingly indispensable. Micro-expression recognition, an invaluable non-verbal communication method, has been extensively studied in human-computer interaction, sentiment analysis, and security fields. However, the sensitivity and privacy implications of micro-expression data pose significant challenges for centralized machine learning methods, raising concerns about serious privacy leakage and data sharing. To address these limitations, we investigate a federated learning scheme tailored specifically for this task. Our approach prioritizes user privacy by employing federated optimization techniques, enabling the aggregation of clients' knowledge in an encrypted space without compromising data privacy. By integrating established micro-expression recognition methods into our framework, we demonstrate that our approach not only ensures robust data protection but also maintains high recognition performance comparable to non-privacy-preserving mechanisms. To our knowledge, this marks the first application of federated learning to the micro-expression recognition task.
{"title":"Towards Federated Learning Driving Technology for Privacy-Preserving Micro-Expression Recognition","authors":"Mingpei Wang;Ling Zhou;Xiaohua Huang;Wenming Zheng","doi":"10.26599/TST.2024.9010098","DOIUrl":"https://doi.org/10.26599/TST.2024.9010098","url":null,"abstract":"As mobile devices and sensor technology advance, their role in communication becomes increasingly indispensable. Micro-expression recognition, an invaluable non-verbal communication method, has been extensively studied in human-computer interaction, sentiment analysis, and security fields. However, the sensitivity and privacy implications of micro-expression data pose significant challenges for centralized machine learning methods, raising concerns about serious privacy leakage and data sharing. To address these limitations, we investigate a federated learning scheme tailored specifically for this task. Our approach prioritizes user privacy by employing federated optimization techniques, enabling the aggregation of clients' knowledge in an encrypted space without compromising data privacy. By integrating established micro-expression recognition methods into our framework, we demonstrate that our approach not only ensures robust data protection but also maintains high recognition performance comparable to non-privacy-preserving mechanisms. To our knowledge, this marks the first application of federated learning to the micro-expression recognition task.","PeriodicalId":48690,"journal":{"name":"Tsinghua Science and Technology","volume":"30 5","pages":"2169-2183"},"PeriodicalIF":6.6,"publicationDate":"2025-04-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10979783","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143888331","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"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 : 2025-04-29DOI: 10.26599/TST.2024.9010152
Xinwei Cao;Xujin Pu;Cheng Hua;Bolin Liao;Ameer Hamza Khan
Neural dynamics is a powerful tool to solve online optimization problems and has been used in many applications. However, some problems cannot be modelled as a single objective optimization and neural dynamics method does not apply. This paper proposes the first neural dynamics model to solve bi-objective constrained quadratic program, which opens the avenue to extend the power of neural dynamics to multi-objective optimization. We rigorously prove that the designed neural dynamics is globally convergent and it converges to the optimal solution of the bi-objective optimization in Pareto sense. Illustrative examples on bi-objective geometric optimization are used to verify the correctness of the proposed method. The developed model is also tested in scientific computing with data from real industrial data with demonstrated superior to rival schemes.
{"title":"Neural Dynamics for Constrained Bi-Objective Quadratic Programming with Applications to Scientific Computing","authors":"Xinwei Cao;Xujin Pu;Cheng Hua;Bolin Liao;Ameer Hamza Khan","doi":"10.26599/TST.2024.9010152","DOIUrl":"https://doi.org/10.26599/TST.2024.9010152","url":null,"abstract":"Neural dynamics is a powerful tool to solve online optimization problems and has been used in many applications. However, some problems cannot be modelled as a single objective optimization and neural dynamics method does not apply. This paper proposes the first neural dynamics model to solve bi-objective constrained quadratic program, which opens the avenue to extend the power of neural dynamics to multi-objective optimization. We rigorously prove that the designed neural dynamics is globally convergent and it converges to the optimal solution of the bi-objective optimization in Pareto sense. Illustrative examples on bi-objective geometric optimization are used to verify the correctness of the proposed method. The developed model is also tested in scientific computing with data from real industrial data with demonstrated superior to rival schemes.","PeriodicalId":48690,"journal":{"name":"Tsinghua Science and Technology","volume":"30 5","pages":"2014-2028"},"PeriodicalIF":6.6,"publicationDate":"2025-04-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10979781","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143888332","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"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}
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：