The frequent missing values in radar-derived time-series tracks of aerial targets(RTT-AT)lead to significant challenges in subsequent data-driven tasks.However,the majority of imputation research focuses on random mis...The frequent missing values in radar-derived time-series tracks of aerial targets(RTT-AT)lead to significant challenges in subsequent data-driven tasks.However,the majority of imputation research focuses on random missing(RM)that differs significantly from common missing patterns of RTT-AT.The method for solving the RM may experience performance degradation or failure when applied to RTT-AT imputation.Conventional autoregressive deep learning methods are prone to error accumulation and long-term dependency loss.In this paper,a non-autoregressive imputation model that addresses the issue of missing value imputation for two common missing patterns in RTT-AT is proposed.Our model consists of two probabilistic sparse diagonal masking self-attention(PSDMSA)units and a weight fusion unit.It learns missing values by combining the representations outputted by the two units,aiming to minimize the difference between the missing values and their actual values.The PSDMSA units effectively capture temporal dependencies and attribute correlations between time steps,improving imputation quality.The weight fusion unit automatically updates the weights of the output representations from the two units to obtain a more accurate final representation.The experimental results indicate that,despite varying missing rates in the two missing patterns,our model consistently outperforms other methods in imputation performance and exhibits a low frequency of deviations in estimates for specific missing entries.Compared to the state-of-the-art autoregressive deep learning imputation model Bidirectional Recurrent Imputation for Time Series(BRITS),our proposed model reduces mean absolute error(MAE)by 31%~50%.Additionally,the model attains a training speed that is 4 to 8 times faster when compared to both BRITS and a standard Transformer model when trained on the same dataset.Finally,the findings from the ablation experiments demonstrate that the PSDMSA,the weight fusion unit,cascade network design,and imputation loss enhance imputation performance and confirm the efficacy of our design.展开更多
Determining trip purpose is an important link to explore travel rules. In this paper,we takea utomobile users in urban areas as the research object,combine unsupervised learning and supervised learningm ethods to anal...Determining trip purpose is an important link to explore travel rules. In this paper,we takea utomobile users in urban areas as the research object,combine unsupervised learning and supervised learningm ethods to analyze their travel characteristics,and focus on the classification and prediction of automobileu sers’trip purposes. However,previous studies on trip purposes mainly focused on questionnaires and GPSd ata,which cannot well reflect the characteristics of automobile travel. In order to avoid the multi-dayb ehavior variability and unobservable heterogeneity of individual characteristics ignored in traditional traffic questionnaires,traffic monitoring data from the Northern District of Qingdao are used,and the K-meansc lustering method is applied to estimate the trip purposes of automobile users. Then,Adaptive Boosting(AdaBoost)and Random Forest(RF)methods are used to classify and predict trip purposes. Finally,ther esult shows:(1)the purpose of automobile users can be mainly divided into four clusters,which includeC ommuting trips,Flexible life demand travel in daytime,Evening entertainment and leisure shopping,andT axi-based trips for the first three types of purposes,respectively;(2)the Random Forest method performss ignificantly better than AdaBoost in trip purpose prediction for higher accuracy;(3)the average predictiona ccuracy of Random Forest under hyper-parameters optimization reaches96.25%,which proves the feasibilitya nd rationality of the above clustering results.展开更多
Background:Intrahepatic cholangiocarcinoma(ICC)is a highly metastatic cancer.^(18)F-fluorodeoxyglucose positron emission tomography/computed tomography(^(18)F-FDG PET/CT)enables sensitive tumor and metastasis detectio...Background:Intrahepatic cholangiocarcinoma(ICC)is a highly metastatic cancer.^(18)F-fluorodeoxyglucose positron emission tomography/computed tomography(^(18)F-FDG PET/CT)enables sensitive tumor and metastasis detection.Our aim is to evaluate the influence of pre-treatment PET/CT on the N-and M-staging and subsequent clinical management in ICC patients.Methods:Between August 2010 and August 2018,660 consecutive ICC patients,without prior anti-tumor treatments nor other malignancies,were enrolled.The diagnostic performance of PET/CT on the N-and M-staging was compared with conventional imaging,and the preoperative staging accuracy and treatment re-allocation by PET/CT were retrospectively calculated.Survival difference was compared between patients receiving PET/CT or not after propensity score matching.Results:Patients were divided into group A(n=291)and group B(n=369)according to whether PET/CT was performed.Among 291 patients with both PET/CT and conventional imaging for staging in group A,PET/CT showed significantly higher sensitivity(83.0%vs.70.5%,P=0.001),specificity(88.3%vs.74.9%,P<0.001)and accuracy(86.3%vs.73.2%,P<0.001)than conventional imaging in diagnosing regional lymph node metastasis,as well as higher sensitivity(87.8%vs.67.6%,P<0.001)and accuracy(93.5%vs.89.3%,P=0.023)in diagnosing distant metastasis.Overall,PET/CT improved the accuracy of preoperative staging from 60.1%to 71.8%(P<0.001),and modified clinical treatment strategy in 5.8%(17/291)of ICC patients,with unique roles in different tumor-node-metastasis(TNM)stages.High tumor-to-non-tumor ratio(TNR)predicted poor overall survival[hazard ratio(HR)=2.17;95%confidence interval(CI):1.49-3.15;P<0.001].Furthermore,patients performing PET/CT had longer overall survival compared with those without PET/CT(HR=0.74;95%CI:0.58-0.93;P=0.011)after propensity score matching.Conclusions:PET/CT was valuable for diagnosing regional lymph node metastasis and distant metastasis in ICC patients,and facilitated accurate tumor staging and optimal treatment allocation.展开更多
Wide-area high-performance computing is widely used for large-scale parallel computing applications owing to its high computing and storage resources.However,the geographical distribution of computing and storage reso...Wide-area high-performance computing is widely used for large-scale parallel computing applications owing to its high computing and storage resources.However,the geographical distribution of computing and storage resources makes efficient task distribution and data placement more challenging.To achieve a higher system performance,this study proposes a two-level global collaborative scheduling strategy for wide-area high-performance computing environments.The collaborative scheduling strategy integrates lightweight solution selection,redundant data placement and task stealing mechanisms,optimizing task distribution and data placement to achieve efficient computing in wide-area environments.The experimental results indicate that compared with the state-of-the-art collaborative scheduling algorithm HPS+,the proposed scheduling strategy reduces the makespan by 23.24%,improves computing and storage resource utilization by 8.28%and 21.73%respectively,and achieves similar global data migration costs.展开更多
With the expansion of cities and the emergence of various urban problems,urban underground space has been developed as a solution.In China’s urban transition context,there is a need for the development of underground...With the expansion of cities and the emergence of various urban problems,urban underground space has been developed as a solution.In China’s urban transition context,there is a need for the development of underground space in urban built-up areas.In this casestudy of the central city of Nanjing,we used spatial analysis and statistical methods to characterize the underground space use of urban built-up areas from a dynamic spatiotemporal perspective.We first analyzed the relationship between the population distribution and the underground space use of the central city of Nanjing based on a Baidu heat map,which can reflect the real-time population distribution,and then,we explored the spatiotemporal characteristics and spatial structure of the underground space use in urban built-up areas.The analysis results provide a reference for planning to improve and optimize the layout of underground space in the central city of Nanjing and,more generally,for the stock-type planning of underground space in urban built-up areas.展开更多
基金supported by Graduate Funded Project(No.JY2022A017).
文摘The frequent missing values in radar-derived time-series tracks of aerial targets(RTT-AT)lead to significant challenges in subsequent data-driven tasks.However,the majority of imputation research focuses on random missing(RM)that differs significantly from common missing patterns of RTT-AT.The method for solving the RM may experience performance degradation or failure when applied to RTT-AT imputation.Conventional autoregressive deep learning methods are prone to error accumulation and long-term dependency loss.In this paper,a non-autoregressive imputation model that addresses the issue of missing value imputation for two common missing patterns in RTT-AT is proposed.Our model consists of two probabilistic sparse diagonal masking self-attention(PSDMSA)units and a weight fusion unit.It learns missing values by combining the representations outputted by the two units,aiming to minimize the difference between the missing values and their actual values.The PSDMSA units effectively capture temporal dependencies and attribute correlations between time steps,improving imputation quality.The weight fusion unit automatically updates the weights of the output representations from the two units to obtain a more accurate final representation.The experimental results indicate that,despite varying missing rates in the two missing patterns,our model consistently outperforms other methods in imputation performance and exhibits a low frequency of deviations in estimates for specific missing entries.Compared to the state-of-the-art autoregressive deep learning imputation model Bidirectional Recurrent Imputation for Time Series(BRITS),our proposed model reduces mean absolute error(MAE)by 31%~50%.Additionally,the model attains a training speed that is 4 to 8 times faster when compared to both BRITS and a standard Transformer model when trained on the same dataset.Finally,the findings from the ablation experiments demonstrate that the PSDMSA,the weight fusion unit,cascade network design,and imputation loss enhance imputation performance and confirm the efficacy of our design.
基金Sponsored by the National Key R&D Program of China(Grant No.2020YFB1600500)the National Natural Science Foundation of China(GrantN o.52272319)。
文摘Determining trip purpose is an important link to explore travel rules. In this paper,we takea utomobile users in urban areas as the research object,combine unsupervised learning and supervised learningm ethods to analyze their travel characteristics,and focus on the classification and prediction of automobileu sers’trip purposes. However,previous studies on trip purposes mainly focused on questionnaires and GPSd ata,which cannot well reflect the characteristics of automobile travel. In order to avoid the multi-dayb ehavior variability and unobservable heterogeneity of individual characteristics ignored in traditional traffic questionnaires,traffic monitoring data from the Northern District of Qingdao are used,and the K-meansc lustering method is applied to estimate the trip purposes of automobile users. Then,Adaptive Boosting(AdaBoost)and Random Forest(RF)methods are used to classify and predict trip purposes. Finally,ther esult shows:(1)the purpose of automobile users can be mainly divided into four clusters,which includeC ommuting trips,Flexible life demand travel in daytime,Evening entertainment and leisure shopping,andT axi-based trips for the first three types of purposes,respectively;(2)the Random Forest method performss ignificantly better than AdaBoost in trip purpose prediction for higher accuracy;(3)the average predictiona ccuracy of Random Forest under hyper-parameters optimization reaches96.25%,which proves the feasibilitya nd rationality of the above clustering results.
基金supported by project grants from the National Natural Science Foundation of China(Nos.91859105,and 81961128025)the Program of Shanghai Academic Research Leader(No.19XD1420700)+1 种基金the Science and Technology Commission of Shanghai Municipality(No.20JC1418900)the Shanghai Municipal Key Clinical Specialty.
文摘Background:Intrahepatic cholangiocarcinoma(ICC)is a highly metastatic cancer.^(18)F-fluorodeoxyglucose positron emission tomography/computed tomography(^(18)F-FDG PET/CT)enables sensitive tumor and metastasis detection.Our aim is to evaluate the influence of pre-treatment PET/CT on the N-and M-staging and subsequent clinical management in ICC patients.Methods:Between August 2010 and August 2018,660 consecutive ICC patients,without prior anti-tumor treatments nor other malignancies,were enrolled.The diagnostic performance of PET/CT on the N-and M-staging was compared with conventional imaging,and the preoperative staging accuracy and treatment re-allocation by PET/CT were retrospectively calculated.Survival difference was compared between patients receiving PET/CT or not after propensity score matching.Results:Patients were divided into group A(n=291)and group B(n=369)according to whether PET/CT was performed.Among 291 patients with both PET/CT and conventional imaging for staging in group A,PET/CT showed significantly higher sensitivity(83.0%vs.70.5%,P=0.001),specificity(88.3%vs.74.9%,P<0.001)and accuracy(86.3%vs.73.2%,P<0.001)than conventional imaging in diagnosing regional lymph node metastasis,as well as higher sensitivity(87.8%vs.67.6%,P<0.001)and accuracy(93.5%vs.89.3%,P=0.023)in diagnosing distant metastasis.Overall,PET/CT improved the accuracy of preoperative staging from 60.1%to 71.8%(P<0.001),and modified clinical treatment strategy in 5.8%(17/291)of ICC patients,with unique roles in different tumor-node-metastasis(TNM)stages.High tumor-to-non-tumor ratio(TNR)predicted poor overall survival[hazard ratio(HR)=2.17;95%confidence interval(CI):1.49-3.15;P<0.001].Furthermore,patients performing PET/CT had longer overall survival compared with those without PET/CT(HR=0.74;95%CI:0.58-0.93;P=0.011)after propensity score matching.Conclusions:PET/CT was valuable for diagnosing regional lymph node metastasis and distant metastasis in ICC patients,and facilitated accurate tumor staging and optimal treatment allocation.
基金This work was supported by the National key R&D Program of China(2018YFB0203901)the National Natural Science Foundation of China under(Grant No.61772053)the fund of the State Key Laboratory of Software Development Environment(SKLSDE-2020ZX15).
文摘Wide-area high-performance computing is widely used for large-scale parallel computing applications owing to its high computing and storage resources.However,the geographical distribution of computing and storage resources makes efficient task distribution and data placement more challenging.To achieve a higher system performance,this study proposes a two-level global collaborative scheduling strategy for wide-area high-performance computing environments.The collaborative scheduling strategy integrates lightweight solution selection,redundant data placement and task stealing mechanisms,optimizing task distribution and data placement to achieve efficient computing in wide-area environments.The experimental results indicate that compared with the state-of-the-art collaborative scheduling algorithm HPS+,the proposed scheduling strategy reduces the makespan by 23.24%,improves computing and storage resource utilization by 8.28%and 21.73%respectively,and achieves similar global data migration costs.
基金the support of the National Natural Science of China(Grant No.51878660)the National Natural Science of China(Grant No.51608527)the Natural Science of Jiangsu Province(Grant No.BK20191330).
文摘With the expansion of cities and the emergence of various urban problems,urban underground space has been developed as a solution.In China’s urban transition context,there is a need for the development of underground space in urban built-up areas.In this casestudy of the central city of Nanjing,we used spatial analysis and statistical methods to characterize the underground space use of urban built-up areas from a dynamic spatiotemporal perspective.We first analyzed the relationship between the population distribution and the underground space use of the central city of Nanjing based on a Baidu heat map,which can reflect the real-time population distribution,and then,we explored the spatiotemporal characteristics and spatial structure of the underground space use in urban built-up areas.The analysis results provide a reference for planning to improve and optimize the layout of underground space in the central city of Nanjing and,more generally,for the stock-type planning of underground space in urban built-up areas.