The formation control of unmanned aerial vehicle (UAV) swarms is of significant importance in various fields such as transportation, emergency management, and environmental monitoring. However, the complex dynamics, nonlinearity, uncertainty, and interaction among agents make it a challenging problem. In this paper, we propose a distributed robust control strategy that uses only local information of UAVs to improve the stability and robustness of the formation system in uncertain environments. We establish a nominal control strategy based on position relations and a semi-definite programming model to obtain control gains. Additionally, we propose a robust control strategy under the rotation set Ω to address the noise and disturbance in the system, ensuring that even when the rotation angles of the UAVs change, they still form a stable formation. Finally, we extend the proposed strategy to a quadrotor UAV system with high-order kinematic models and conduct simulation experiments to validate its effectiveness in resisting uncertain disturbances and achieving formation control.
{"title":"Distributed Robust UAVs Formation Control Based on Semidefinite Programming","authors":"Peiyu Zhang;Jianshan Zhou;Daxin Tian;Xuting Duan;Dezong Zhao;Kan Guo","doi":"10.26599/TST.2023.9010125","DOIUrl":"https://doi.org/10.26599/TST.2023.9010125","url":null,"abstract":"The formation control of unmanned aerial vehicle (UAV) swarms is of significant importance in various fields such as transportation, emergency management, and environmental monitoring. However, the complex dynamics, nonlinearity, uncertainty, and interaction among agents make it a challenging problem. In this paper, we propose a distributed robust control strategy that uses only local information of UAVs to improve the stability and robustness of the formation system in uncertain environments. We establish a nominal control strategy based on position relations and a semi-definite programming model to obtain control gains. Additionally, we propose a robust control strategy under the rotation set Ω to address the noise and disturbance in the system, ensuring that even when the rotation angles of the UAVs change, they still form a stable formation. Finally, we extend the proposed strategy to a quadrotor UAV system with high-order kinematic models and conduct simulation experiments to validate its effectiveness in resisting uncertain disturbances and achieving formation control.","PeriodicalId":48690,"journal":{"name":"Tsinghua Science and Technology","volume":null,"pages":null},"PeriodicalIF":6.6,"publicationDate":"2024-03-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10517976","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140820231","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 : 2024-03-02DOI: 10.26599/TST.2023.9010144
Ting Cheng;Luqing Liu;Zhongzhu Li;Siyu Heng
Dwell scheduling is a key for phased array radar to realize multi-function and it becomes especially challenging in complex tactical situations. In this manuscript, a real-time radar dwell scheduling algorithm based on a unified pulse interleaving framework is proposed. A unified pulse interleaving framework that can realize pulse interleaving analysis for phased array radars with different receiving modes is put forward, which greatly improves the time utilization of the system. Based on above framework, a real-time two-stage approach is proposed to solve the optimization problem of dwell scheduling. The importance and urgency criteria are guaranteed by the first pre-schedule stage, and the desired execution time criterion is improved at the second stage with the modified particle swarm optimization (PSO). Simulation results demonstrate that the proposed algorithm has better comprehensive scheduling performance than up-to-date algorithms that consider the pulse interleaving technique for both single beam and multiple beams receiving modes. Besides, the proposed algorithm can realize dwell scheduling in realtime.
{"title":"Real-Time Dwell Scheduling Based on a Unified Pulse Interleaving Framework for Phased Array Radar","authors":"Ting Cheng;Luqing Liu;Zhongzhu Li;Siyu Heng","doi":"10.26599/TST.2023.9010144","DOIUrl":"https://doi.org/10.26599/TST.2023.9010144","url":null,"abstract":"Dwell scheduling is a key for phased array radar to realize multi-function and it becomes especially challenging in complex tactical situations. In this manuscript, a real-time radar dwell scheduling algorithm based on a unified pulse interleaving framework is proposed. A unified pulse interleaving framework that can realize pulse interleaving analysis for phased array radars with different receiving modes is put forward, which greatly improves the time utilization of the system. Based on above framework, a real-time two-stage approach is proposed to solve the optimization problem of dwell scheduling. The importance and urgency criteria are guaranteed by the first pre-schedule stage, and the desired execution time criterion is improved at the second stage with the modified particle swarm optimization (PSO). Simulation results demonstrate that the proposed algorithm has better comprehensive scheduling performance than up-to-date algorithms that consider the pulse interleaving technique for both single beam and multiple beams receiving modes. Besides, the proposed algorithm can realize dwell scheduling in realtime.","PeriodicalId":48690,"journal":{"name":"Tsinghua Science and Technology","volume":null,"pages":null},"PeriodicalIF":6.6,"publicationDate":"2024-03-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10517979","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140820232","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}
Breast mass identification is of great significance for early screening of breast cancer, while the existing detection methods have high missed and misdiagnosis rate for small masses. We propose a small target breast mass detection network named Residual asymmetric dilated convolution-Cross layer attention-Mean standard deviation adaptive selection-You Only Look Once (RCM-YOLO), which improves the identifiability of small masses by increasing the resolution of feature maps, adopts residual asymmetric dilated convolution to expand the receptive field and optimize the amount of parameters, and proposes the cross-layer attention that transfers the deep semantic information to the shallow layer as auxiliary information to obtain key feature locations. In the training process, we propose an adaptive positive sample selection algorithm to automatically select positive samples, which considers the statistical features of the intersection over union sets to ensure the validity of the training set and the detection accuracy of the model. To verify the performance of our model, we used public datasets to carry out the experiments. The results showed that the mean Average Precision (mAP) of RCM-YOLO reached 90.34%, compared with YOLOv5, the missed detection rate for small masses of RCM-YOLO was reduced to 11%, and the single detection time was reduced to 28 ms. The detection accuracy and speed can be effectively improved by strengthening the feature expression of small masses and the relationship between features. Our method can help doctors in batch screening of breast images, and significantly promote the detection rate of small masses and reduce misdiagnosis.
{"title":"Detection and Diagnosis of Small Target Breast Masses Based on Convolutional Neural Networks","authors":"Ling Tan;Ying Liang;Jingming Xia;Hui Wu;Jining Zhu","doi":"10.26599/TST.2023.9010126","DOIUrl":"https://doi.org/10.26599/TST.2023.9010126","url":null,"abstract":"Breast mass identification is of great significance for early screening of breast cancer, while the existing detection methods have high missed and misdiagnosis rate for small masses. We propose a small target breast mass detection network named Residual asymmetric dilated convolution-Cross layer attention-Mean standard deviation adaptive selection-You Only Look Once (RCM-YOLO), which improves the identifiability of small masses by increasing the resolution of feature maps, adopts residual asymmetric dilated convolution to expand the receptive field and optimize the amount of parameters, and proposes the cross-layer attention that transfers the deep semantic information to the shallow layer as auxiliary information to obtain key feature locations. In the training process, we propose an adaptive positive sample selection algorithm to automatically select positive samples, which considers the statistical features of the intersection over union sets to ensure the validity of the training set and the detection accuracy of the model. To verify the performance of our model, we used public datasets to carry out the experiments. The results showed that the mean Average Precision (mAP) of RCM-YOLO reached 90.34%, compared with YOLOv5, the missed detection rate for small masses of RCM-YOLO was reduced to 11%, and the single detection time was reduced to 28 ms. The detection accuracy and speed can be effectively improved by strengthening the feature expression of small masses and the relationship between features. Our method can help doctors in batch screening of breast images, and significantly promote the detection rate of small masses and reduce misdiagnosis.","PeriodicalId":48690,"journal":{"name":"Tsinghua Science and Technology","volume":null,"pages":null},"PeriodicalIF":6.6,"publicationDate":"2024-03-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10517921","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140820225","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 : 2024-03-02DOI: 10.26599/TST.2023.9010076
Feng Wang;Qi He;Shicheng Li
Combinatorial Optimization Problems (COPs) are a class of optimization problems that are commonly encountered in industrial production and everyday life. Over the last few decades, traditional algorithms, such as exact algorithms, approximate algorithms, and heuristic algorithms, have been proposed to solve COPs. However, as COPs in the real world become more complex, traditional algorithms struggle to generate optimal solutions in a limited amount of time. Since Deep Neural Networks (DNNs) are not heavily dependent on expert knowledge and are adequately flexible for generalization to various COPs, several DNN-based algorithms have been proposed in the last ten years for solving COPs. Herein, we categorize these algorithms into four classes and provide a brief overview of their applications in real-world problems.
{"title":"Solving Combinatorial Optimization Problems with Deep Neural Network: A Survey","authors":"Feng Wang;Qi He;Shicheng Li","doi":"10.26599/TST.2023.9010076","DOIUrl":"https://doi.org/10.26599/TST.2023.9010076","url":null,"abstract":"Combinatorial Optimization Problems (COPs) are a class of optimization problems that are commonly encountered in industrial production and everyday life. Over the last few decades, traditional algorithms, such as exact algorithms, approximate algorithms, and heuristic algorithms, have been proposed to solve COPs. However, as COPs in the real world become more complex, traditional algorithms struggle to generate optimal solutions in a limited amount of time. Since Deep Neural Networks (DNNs) are not heavily dependent on expert knowledge and are adequately flexible for generalization to various COPs, several DNN-based algorithms have been proposed in the last ten years for solving COPs. Herein, we categorize these algorithms into four classes and provide a brief overview of their applications in real-world problems.","PeriodicalId":48690,"journal":{"name":"Tsinghua Science and Technology","volume":null,"pages":null},"PeriodicalIF":6.6,"publicationDate":"2024-03-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10517929","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140820180","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 : 2024-03-02DOI: 10.26599/TST.2023.9010098
Wei Li;Xiangfang Yan;Ying Huang
With the advancement of the manufacturing industry, the investigation of the shop floor scheduling problem has gained increasing importance. The Job shop Scheduling Problem (JSP), as a fundamental scheduling problem, holds considerable theoretical research value. However, finding a satisfactory solution within a given time is difficult due to the NP-hard nature of the JSP. A co-operative-guided ant colony optimization algorithm with knowledge learning (namely KLCACO) is proposed to address this difficulty. This algorithm integrates a data-based swarm intelligence optimization algorithm with model-based JSP schedule knowledge. A solution construction scheme based on scheduling knowledge learning is proposed for KLCACO. The problem model and algorithm data are fused by merging scheduling and planning knowledge with individual scheme construction to enhance the quality of the generated individual solutions. A pheromone guidance mechanism, which is based on a collaborative machine strategy, is used to simplify information learning and the problem space by collaborating with different machine processing orders. Additionally, the KLCACO algorithm utilizes the classical neighborhood structure to optimize the solution, expanding the search space of the algorithm and accelerating its convergence. The KLCACO algorithm is compared with other high-performance intelligent optimization algorithms on four public benchmark datasets, comprising 48 benchmark test cases in total. The effectiveness of the proposed algorithm in addressing JSPs is validated, demonstrating the feasibility of the KLCACO algorithm for knowledge and data fusion in complex combinatorial optimization problems.
{"title":"Cooperative-Guided Ant Colony Optimization with Knowledge Learning for Job Shop Scheduling Problem","authors":"Wei Li;Xiangfang Yan;Ying Huang","doi":"10.26599/TST.2023.9010098","DOIUrl":"https://doi.org/10.26599/TST.2023.9010098","url":null,"abstract":"With the advancement of the manufacturing industry, the investigation of the shop floor scheduling problem has gained increasing importance. The Job shop Scheduling Problem (JSP), as a fundamental scheduling problem, holds considerable theoretical research value. However, finding a satisfactory solution within a given time is difficult due to the NP-hard nature of the JSP. A co-operative-guided ant colony optimization algorithm with knowledge learning (namely KLCACO) is proposed to address this difficulty. This algorithm integrates a data-based swarm intelligence optimization algorithm with model-based JSP schedule knowledge. A solution construction scheme based on scheduling knowledge learning is proposed for KLCACO. The problem model and algorithm data are fused by merging scheduling and planning knowledge with individual scheme construction to enhance the quality of the generated individual solutions. A pheromone guidance mechanism, which is based on a collaborative machine strategy, is used to simplify information learning and the problem space by collaborating with different machine processing orders. Additionally, the KLCACO algorithm utilizes the classical neighborhood structure to optimize the solution, expanding the search space of the algorithm and accelerating its convergence. The KLCACO algorithm is compared with other high-performance intelligent optimization algorithms on four public benchmark datasets, comprising 48 benchmark test cases in total. The effectiveness of the proposed algorithm in addressing JSPs is validated, demonstrating the feasibility of the KLCACO algorithm for knowledge and data fusion in complex combinatorial optimization problems.","PeriodicalId":48690,"journal":{"name":"Tsinghua Science and Technology","volume":null,"pages":null},"PeriodicalIF":6.6,"publicationDate":"2024-03-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10517930","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140820181","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 : 2024-03-02DOI: 10.26599/TST.2023.9010123
Guoting Zhang;Yonghao Du;Xiaobin Zhu;Xiaolu Liu
Multimodal multi-objective optimization problems (MMOPs) contain multiple equivalent Pareto subsets (PSs) corresponding to a single Pareto front (PF), resulting in difficulty in maintaining promising diversities in both objective and decision spaces to find these PSs. Widely used to solve MMOPs, evolutionary algorithms mainly consist of evolutionary operators that generate new solutions and fitness evaluations of the solutions. To enhance performance in solving MMOPs, this paper proposes a multimodal multi-objective optimization evolutionary algorithm based on a hybrid operator and strengthened diversity improving. Specifically, a hybrid operator mechanism is devised to ensure the exploration of the decision space in the early stage and approximation to the optima in the latter stage. Moreover, an elitist-assisted differential evolution mechanism is designed for the early exploration stage. In addition, a new fitness function is proposed and used in environmental and mating selections to simultaneously evaluate diversities for PF and PSs. Experimental studies on 11 widely used benchmark instances from a test suite verify the superiority or at least competitiveness of the proposed methods compared to five state-of-the-art algorithms tailored for MMOPs.
{"title":"Hybrid Operator and Strengthened Diversity Improving for Multimodal Multi-Objective Optimization","authors":"Guoting Zhang;Yonghao Du;Xiaobin Zhu;Xiaolu Liu","doi":"10.26599/TST.2023.9010123","DOIUrl":"https://doi.org/10.26599/TST.2023.9010123","url":null,"abstract":"Multimodal multi-objective optimization problems (MMOPs) contain multiple equivalent Pareto subsets (PSs) corresponding to a single Pareto front (PF), resulting in difficulty in maintaining promising diversities in both objective and decision spaces to find these PSs. Widely used to solve MMOPs, evolutionary algorithms mainly consist of evolutionary operators that generate new solutions and fitness evaluations of the solutions. To enhance performance in solving MMOPs, this paper proposes a multimodal multi-objective optimization evolutionary algorithm based on a hybrid operator and strengthened diversity improving. Specifically, a hybrid operator mechanism is devised to ensure the exploration of the decision space in the early stage and approximation to the optima in the latter stage. Moreover, an elitist-assisted differential evolution mechanism is designed for the early exploration stage. In addition, a new fitness function is proposed and used in environmental and mating selections to simultaneously evaluate diversities for PF and PSs. Experimental studies on 11 widely used benchmark instances from a test suite verify the superiority or at least competitiveness of the proposed methods compared to five state-of-the-art algorithms tailored for MMOPs.","PeriodicalId":48690,"journal":{"name":"Tsinghua Science and Technology","volume":null,"pages":null},"PeriodicalIF":6.6,"publicationDate":"2024-03-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10517919","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140820227","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 : 2024-03-02DOI: 10.26599/TST.2023.9010103
Xi Chen;Quan Cheng
Acute complication prediction model is of great importance for the overall reduction of premature death in chronic diseases. The CLSTM-BPR proposed in this paper aims to improve the accuracy, interpretability, and generalizability of the existing disease prediction models. Firstly, through its complex neural network structure, CLSTM-BPR considers both disease commonality and patient characteristics in the prediction process. Secondly, by splicing the time series prediction algorithm and classifier, the judgment basis is given along with the prediction results. Finally, this model introduces the pairwise algorithm Bayesian Personalized Ranking (BPR) into the medical field for the first time, and achieves a good result in the diagnosis of six acute complications. Experiments on the Medical Information Mart for Intensive Care IV (MIMIC-IV) dataset show that the average Mean Absolute Error (MAE) of biomarker value prediction of the CLSTM-BPR model is 0.26, and the average accuracy (ACC) of the CLSTM-BPR model for acute complication diagnosis is 92.5%. Comparison experiments and ablation experiments further demonstrate the reliability of CLSTM-BPR in the prediction of acute complication, which is an advancement of current disease prediction tools.
{"title":"Acute Complication Prediction and Diagnosis Model CLSTM-BPR: A Fusion Method of Time Series Deep Learning and Bayesian Personalized Ranking","authors":"Xi Chen;Quan Cheng","doi":"10.26599/TST.2023.9010103","DOIUrl":"https://doi.org/10.26599/TST.2023.9010103","url":null,"abstract":"Acute complication prediction model is of great importance for the overall reduction of premature death in chronic diseases. The CLSTM-BPR proposed in this paper aims to improve the accuracy, interpretability, and generalizability of the existing disease prediction models. Firstly, through its complex neural network structure, CLSTM-BPR considers both disease commonality and patient characteristics in the prediction process. Secondly, by splicing the time series prediction algorithm and classifier, the judgment basis is given along with the prediction results. Finally, this model introduces the pairwise algorithm Bayesian Personalized Ranking (BPR) into the medical field for the first time, and achieves a good result in the diagnosis of six acute complications. Experiments on the Medical Information Mart for Intensive Care IV (MIMIC-IV) dataset show that the average Mean Absolute Error (MAE) of biomarker value prediction of the CLSTM-BPR model is 0.26, and the average accuracy (ACC) of the CLSTM-BPR model for acute complication diagnosis is 92.5%. Comparison experiments and ablation experiments further demonstrate the reliability of CLSTM-BPR in the prediction of acute complication, which is an advancement of current disease prediction tools.","PeriodicalId":48690,"journal":{"name":"Tsinghua Science and Technology","volume":null,"pages":null},"PeriodicalIF":6.6,"publicationDate":"2024-03-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10517923","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140820428","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}
Reinforcement learning holds promise in enabling robotic tasks as it can learn optimal policies via trial and error. However, the practical deployment of reinforcement learning usually requires human intervention to provide episodic resets when a failure occurs. Since manual resets are generally unavailable in autonomous robots, we propose a reset-free reinforcement learning algorithm based on multi-state recovery and failure prevention to avoid failure-induced resets. The multi-state recovery provides robots with the capability of recovering from failures by self-correcting its behavior in the problematic state and, more importantly, deciding which previous state is the best to return to for efficient re-learning. The failure prevention reduces potential failures by predicting and excluding possible unsafe actions in specific states. Both simulations and real-world experiments are used to validate our algorithm with the results showing a significant reduction in the number of resets and failures during the learning.
{"title":"Reset-Free Reinforcement Learning via Multi-State Recovery and Failure Prevention for Autonomous Robots","authors":"Xu Zhou;Benlian Xu;Zhengqiang Jiang;Jun Li;Brett Nener","doi":"10.26599/TST.2023.9010117","DOIUrl":"https://doi.org/10.26599/TST.2023.9010117","url":null,"abstract":"Reinforcement learning holds promise in enabling robotic tasks as it can learn optimal policies via trial and error. However, the practical deployment of reinforcement learning usually requires human intervention to provide episodic resets when a failure occurs. Since manual resets are generally unavailable in autonomous robots, we propose a reset-free reinforcement learning algorithm based on multi-state recovery and failure prevention to avoid failure-induced resets. The multi-state recovery provides robots with the capability of recovering from failures by self-correcting its behavior in the problematic state and, more importantly, deciding which previous state is the best to return to for efficient re-learning. The failure prevention reduces potential failures by predicting and excluding possible unsafe actions in specific states. Both simulations and real-world experiments are used to validate our algorithm with the results showing a significant reduction in the number of resets and failures during the learning.","PeriodicalId":48690,"journal":{"name":"Tsinghua Science and Technology","volume":null,"pages":null},"PeriodicalIF":6.6,"publicationDate":"2024-03-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10517924","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140820430","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}
Currently, most existing inductive relation prediction approaches are based on subgraph structures, with subgraph features extracted using graph neural networks to predict relations. However, subgraphs may contain disconnected regions, which usually represent different semantic ranges. Because not all semantic information about the regions is helpful in relation prediction, we propose a relation prediction model based on a disentangled subgraph structure and implement a feature updating approach based on relevant semantic aggregation. To indirectly achieve the disentangled subgraph structure from a semantic perspective, the mapping of entity features into different semantic spaces and the aggregation of related semantics on each semantic space are updated. The disentangled model can focus on features having higher semantic relevance in the prediction, thus addressing a problem with existing approaches, which ignore the semantic differences in different subgraph structures. Furthermore, using a gated recurrent neural network, this model enhances the features of entities by sorting them by distance and extracting the path information in the subgraphs. Experimentally, it is shown that when there are numerous disconnected regions in the subgraph, our model outperforms existing mainstream models in terms of both Area Under the Curve-Precision-Recall (AUC-PR) and Hits@10. Experiments prove that semantic differences in the knowledge graph can be effectively distinguished and verify the effectiveness of this method.
{"title":"Inductive Relation Prediction by Disentangled Subgraph Structure","authors":"Guiduo Duan;Rui Guo;Wenlong Luo;Guangchun Luo;Tianxi Huang","doi":"10.26599/TST.2023.9010154","DOIUrl":"https://doi.org/10.26599/TST.2023.9010154","url":null,"abstract":"Currently, most existing inductive relation prediction approaches are based on subgraph structures, with subgraph features extracted using graph neural networks to predict relations. However, subgraphs may contain disconnected regions, which usually represent different semantic ranges. Because not all semantic information about the regions is helpful in relation prediction, we propose a relation prediction model based on a disentangled subgraph structure and implement a feature updating approach based on relevant semantic aggregation. To indirectly achieve the disentangled subgraph structure from a semantic perspective, the mapping of entity features into different semantic spaces and the aggregation of related semantics on each semantic space are updated. The disentangled model can focus on features having higher semantic relevance in the prediction, thus addressing a problem with existing approaches, which ignore the semantic differences in different subgraph structures. Furthermore, using a gated recurrent neural network, this model enhances the features of entities by sorting them by distance and extracting the path information in the subgraphs. Experimentally, it is shown that when there are numerous disconnected regions in the subgraph, our model outperforms existing mainstream models in terms of both Area Under the Curve-Precision-Recall (AUC-PR) and Hits@10. Experiments prove that semantic differences in the knowledge graph can be effectively distinguished and verify the effectiveness of this method.","PeriodicalId":48690,"journal":{"name":"Tsinghua Science and Technology","volume":null,"pages":null},"PeriodicalIF":6.6,"publicationDate":"2024-03-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10517981","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140820183","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 : 2024-03-02DOI: 10.26599/TST.2023.9010092
Lan Lan;Guoliang Hu;Rui Li;Tingting Wang;Lingling Jiang;Jiawei Luo;Zhiwei Ji;Yilong Wang
Differences in the imaging subgroups of cerebral small vessel disease (CSVD) need to be further explored. First, we use propensity score matching to obtain balanced datasets. Then random forest (RF) is adopted to classify the subgroups compared with support vector machine (SVM) and extreme gradient boosting (XGBoost), and to select the features. The top 10 important features are included in the stepwise logistic regression, and the odds ratio (OR) and 95% confidence interval (Cl) are obtained. There are 41 290 adult inpatient records diagnosed with CSVD. Accuracy and area under curve (AUC) of RF are close to 0.7, which performs best in classification compared to SVM and XGBoost. OR and 95% Cl of hematocrit for white matter lesions (WMLs), lacunes, microbleeds, atrophy, and enlarged perivascular space (EPVS) are 0.9875 (0.9857–0.9893), 0.9728 (0.9705–0.9752), 0.9782 (0.9740–0.9824), 1.0093 (1.0081–1.0106), and 0.9716 (0.9597–0.9832). OR and 95% Cl of red cell distribution width for WMLs, lacunes, atrophy, and EPVS are 0.9600 (0.9538–0.9662), 0.9630 (0.9559–0.9702), 1.0751 (1.0686–1.0817), and 0.9304 (0.8864–0.9755). OR and 95% Cl of platelet distribution width for WMLs, lacunes, and microbleeds are 1.1796 (1.1636–1.1958), 1.1663 (1.1476–1.1853), and 1.0416 (1.0152–1.0687). This study proposes a new analytical framework to select important clinical markers for CSVD with machine learning based on a common data model, which has low cost, fast speed, large sample size, and continuous data sources.
{"title":"Machine Learning for Selecting Important Clinical Markers of Imaging Subgroups of Cerebral Small Vessel Disease Based on a Common Data Model","authors":"Lan Lan;Guoliang Hu;Rui Li;Tingting Wang;Lingling Jiang;Jiawei Luo;Zhiwei Ji;Yilong Wang","doi":"10.26599/TST.2023.9010092","DOIUrl":"https://doi.org/10.26599/TST.2023.9010092","url":null,"abstract":"Differences in the imaging subgroups of cerebral small vessel disease (CSVD) need to be further explored. First, we use propensity score matching to obtain balanced datasets. Then random forest (RF) is adopted to classify the subgroups compared with support vector machine (SVM) and extreme gradient boosting (XGBoost), and to select the features. The top 10 important features are included in the stepwise logistic regression, and the odds ratio (OR) and 95% confidence interval (Cl) are obtained. There are 41 290 adult inpatient records diagnosed with CSVD. Accuracy and area under curve (AUC) of RF are close to 0.7, which performs best in classification compared to SVM and XGBoost. OR and 95% Cl of hematocrit for white matter lesions (WMLs), lacunes, microbleeds, atrophy, and enlarged perivascular space (EPVS) are 0.9875 (0.9857–0.9893), 0.9728 (0.9705–0.9752), 0.9782 (0.9740–0.9824), 1.0093 (1.0081–1.0106), and 0.9716 (0.9597–0.9832). OR and 95% Cl of red cell distribution width for WMLs, lacunes, atrophy, and EPVS are 0.9600 (0.9538–0.9662), 0.9630 (0.9559–0.9702), 1.0751 (1.0686–1.0817), and 0.9304 (0.8864–0.9755). OR and 95% Cl of platelet distribution width for WMLs, lacunes, and microbleeds are 1.1796 (1.1636–1.1958), 1.1663 (1.1476–1.1853), and 1.0416 (1.0152–1.0687). This study proposes a new analytical framework to select important clinical markers for CSVD with machine learning based on a common data model, which has low cost, fast speed, large sample size, and continuous data sources.","PeriodicalId":48690,"journal":{"name":"Tsinghua Science and Technology","volume":null,"pages":null},"PeriodicalIF":6.6,"publicationDate":"2024-03-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10517920","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140820241","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}