In this paper,we established a class of parallel algorithm for solving low-rank tensor completion problem.The main idea is that N singular value decompositions are implemented in N different processors for each slice ...In this paper,we established a class of parallel algorithm for solving low-rank tensor completion problem.The main idea is that N singular value decompositions are implemented in N different processors for each slice matrix under unfold operator,and then the fold operator is used to form the next iteration tensor such that the computing time can be decreased.In theory,we analyze the global convergence of the algorithm.In numerical experiment,the simulation data and real image inpainting are carried out.Experiment results show the parallel algorithm outperform its original algorithm in CPU times under the same precision.展开更多
Aiming at the problem that infrared small target detection faces low contrast between the background and the target and insufficient noise suppression ability under the complex cloud background,an infrared small targe...Aiming at the problem that infrared small target detection faces low contrast between the background and the target and insufficient noise suppression ability under the complex cloud background,an infrared small target detection method based on the tensor nuclear norm and direction residual weighting was proposed.Based on converting the infrared image into an infrared patch tensor model,from the perspective of the low-rank nature of the background tensor,and taking advantage of the difference in contrast between the background and the target in different directions,we designed a double-neighborhood local contrast based on direction residual weighting method(DNLCDRW)combined with the partial sum of tensor nuclear norm(PSTNN)to achieve effective background suppression and recovery of infrared small targets.Experiments show that the algorithm is effective in suppressing the background and improving the detection ability of the target.展开更多
考虑速度分量的各向异性进行能量估计,得到三维稳态Q-tensor液晶流系统的Liouville型定理,即若u∈L^(q)(R^(3))∩˙H^(1)(R^(3)),u_(i)∈L xi q/q−2 L s xei(R×R^(2))(i=1,2,3),且Q∈H^(2)(R^(3)),其中2/q+1/s≥1/2,1≤s≤∞,2<...考虑速度分量的各向异性进行能量估计,得到三维稳态Q-tensor液晶流系统的Liouville型定理,即若u∈L^(q)(R^(3))∩˙H^(1)(R^(3)),u_(i)∈L xi q/q−2 L s xei(R×R^(2))(i=1,2,3),且Q∈H^(2)(R^(3)),其中2/q+1/s≥1/2,1≤s≤∞,2<q<∞,则该稳态系统只有平凡解.这个结论推广了已有的结果.展开更多
科学的轨道交通出行模式分析是运营决策优化的重要依据。为挖掘城市轨道交通时空流动特征及其影响机理,提出一种基于非负张量分解的OD客流强度时空分布计算方法,采用融合SHAP归因分析的极端梯度提升树(eXtreme Gradient Boosting,XGBoo...科学的轨道交通出行模式分析是运营决策优化的重要依据。为挖掘城市轨道交通时空流动特征及其影响机理,提出一种基于非负张量分解的OD客流强度时空分布计算方法,采用融合SHAP归因分析的极端梯度提升树(eXtreme Gradient Boosting,XGBoost)对各模式OD客流强度进行拟合预测。使用城市轨道交通AFC(automatic fare collection system,AFC)系统数据,从空间、时段以及出行日3个维度构建3阶客流OD张量,采用交替非负最小二乘法(alternating non negative least squares,ANLS)实现非负CP张量分解。基于张量分解结果,从北京轨道交通344个站点连续1周16266966条出行数据中,提取出晨高峰长距离通勤、早高峰中短通勤、平峰休闲中转出行、晚归出行4种出行模式的时、空分布特征。基于可解释性机器学习模型,对各模式OD客流进行预测。结果表明XGBoost与CatBoost、LightGBM、OLS相比更具优势。根据OD起终点站域环境特征,考虑起终点缓冲区内各类兴趣点(point of interest,POI)数量、小区住户数、房价、人口数量、站点偏离距离以及出行距离等指标,构建OD强度关联指标体系,解释各指标对OD客流强度的正负反馈效应。SHAP归因分析说明,居民更倾向于14站以内的中短途出行,并分别得到了就业类POI数目对晨、早通勤客流正向影响,以及餐饮类POI数目对休闲中转出行客流正向影响的临界阈值。该方法可为轨道交通精细化出行引导和客流组织提供数据支撑,优化城市轨道交通供需平衡及服务水平。展开更多
目的探讨抑郁障碍和双相障碍患者脑白质网络节点强度的差异,分析患者不同脑区的结构连接受损情况及其在鉴别中的作用。方法纳入91例基线诊断为抑郁发作的患者,经过≥9年的自然观察随访后,最终确定23例维持抑郁障碍诊断(单相组)和18例维...目的探讨抑郁障碍和双相障碍患者脑白质网络节点强度的差异,分析患者不同脑区的结构连接受损情况及其在鉴别中的作用。方法纳入91例基线诊断为抑郁发作的患者,经过≥9年的自然观察随访后,最终确定23例维持抑郁障碍诊断(单相组)和18例维持双相障碍诊断(双相组)的患者纳入分析。同时纳入30名健康对照者(对照组)。受试者在入组时均接受弥散张量成像扫描,采用确定性纤维追踪技术构建脑白质结构加权网络。比较三组间脑白质网络的节点连接强度差异,进一步采用受试者操作特征(receiver operator characteristic,ROC)曲线评估差异脑区对抑郁障碍和双相障碍鉴别诊断的价值。结果双相组在左前扣带回的节点强度较单相组降低(3.89±0.76 vs.4.74±0.60),在右尾状核(4.94±1.26 vs.3.46±0.99)、右苍白球(1.98±0.67 vs.1.25±0.29)的节点强度较单相组升高(P<0.01,FWE校正)。左前扣带回、右尾状核、右苍白球3个脑区的连接强度联合鉴别抑郁障碍和双相障碍绘制ROC曲线,曲线下面积(area under the curve,AUC)为0.95(95%CI:0.91~0.99;P<0.01),敏感度0.89,特异度0.87。结论脑结构网络的节点强度差异可以作为一个潜在的影像学生物标志物识别抑郁障碍和双相障碍,联合差异脑区的节点强度可以得到更好的识别率。展开更多
基金Supported by National Nature Science Foundation(12371381)Nature Science Foundation of Shanxi(202403021222270)。
文摘In this paper,we established a class of parallel algorithm for solving low-rank tensor completion problem.The main idea is that N singular value decompositions are implemented in N different processors for each slice matrix under unfold operator,and then the fold operator is used to form the next iteration tensor such that the computing time can be decreased.In theory,we analyze the global convergence of the algorithm.In numerical experiment,the simulation data and real image inpainting are carried out.Experiment results show the parallel algorithm outperform its original algorithm in CPU times under the same precision.
基金Supported by the Key Laboratory Fund for Equipment Pre-Research(6142207210202)。
文摘Aiming at the problem that infrared small target detection faces low contrast between the background and the target and insufficient noise suppression ability under the complex cloud background,an infrared small target detection method based on the tensor nuclear norm and direction residual weighting was proposed.Based on converting the infrared image into an infrared patch tensor model,from the perspective of the low-rank nature of the background tensor,and taking advantage of the difference in contrast between the background and the target in different directions,we designed a double-neighborhood local contrast based on direction residual weighting method(DNLCDRW)combined with the partial sum of tensor nuclear norm(PSTNN)to achieve effective background suppression and recovery of infrared small targets.Experiments show that the algorithm is effective in suppressing the background and improving the detection ability of the target.
文摘考虑速度分量的各向异性进行能量估计,得到三维稳态Q-tensor液晶流系统的Liouville型定理,即若u∈L^(q)(R^(3))∩˙H^(1)(R^(3)),u_(i)∈L xi q/q−2 L s xei(R×R^(2))(i=1,2,3),且Q∈H^(2)(R^(3)),其中2/q+1/s≥1/2,1≤s≤∞,2<q<∞,则该稳态系统只有平凡解.这个结论推广了已有的结果.
文摘科学的轨道交通出行模式分析是运营决策优化的重要依据。为挖掘城市轨道交通时空流动特征及其影响机理,提出一种基于非负张量分解的OD客流强度时空分布计算方法,采用融合SHAP归因分析的极端梯度提升树(eXtreme Gradient Boosting,XGBoost)对各模式OD客流强度进行拟合预测。使用城市轨道交通AFC(automatic fare collection system,AFC)系统数据,从空间、时段以及出行日3个维度构建3阶客流OD张量,采用交替非负最小二乘法(alternating non negative least squares,ANLS)实现非负CP张量分解。基于张量分解结果,从北京轨道交通344个站点连续1周16266966条出行数据中,提取出晨高峰长距离通勤、早高峰中短通勤、平峰休闲中转出行、晚归出行4种出行模式的时、空分布特征。基于可解释性机器学习模型,对各模式OD客流进行预测。结果表明XGBoost与CatBoost、LightGBM、OLS相比更具优势。根据OD起终点站域环境特征,考虑起终点缓冲区内各类兴趣点(point of interest,POI)数量、小区住户数、房价、人口数量、站点偏离距离以及出行距离等指标,构建OD强度关联指标体系,解释各指标对OD客流强度的正负反馈效应。SHAP归因分析说明,居民更倾向于14站以内的中短途出行,并分别得到了就业类POI数目对晨、早通勤客流正向影响,以及餐饮类POI数目对休闲中转出行客流正向影响的临界阈值。该方法可为轨道交通精细化出行引导和客流组织提供数据支撑,优化城市轨道交通供需平衡及服务水平。
文摘目的探讨抑郁障碍和双相障碍患者脑白质网络节点强度的差异,分析患者不同脑区的结构连接受损情况及其在鉴别中的作用。方法纳入91例基线诊断为抑郁发作的患者,经过≥9年的自然观察随访后,最终确定23例维持抑郁障碍诊断(单相组)和18例维持双相障碍诊断(双相组)的患者纳入分析。同时纳入30名健康对照者(对照组)。受试者在入组时均接受弥散张量成像扫描,采用确定性纤维追踪技术构建脑白质结构加权网络。比较三组间脑白质网络的节点连接强度差异,进一步采用受试者操作特征(receiver operator characteristic,ROC)曲线评估差异脑区对抑郁障碍和双相障碍鉴别诊断的价值。结果双相组在左前扣带回的节点强度较单相组降低(3.89±0.76 vs.4.74±0.60),在右尾状核(4.94±1.26 vs.3.46±0.99)、右苍白球(1.98±0.67 vs.1.25±0.29)的节点强度较单相组升高(P<0.01,FWE校正)。左前扣带回、右尾状核、右苍白球3个脑区的连接强度联合鉴别抑郁障碍和双相障碍绘制ROC曲线,曲线下面积(area under the curve,AUC)为0.95(95%CI:0.91~0.99;P<0.01),敏感度0.89,特异度0.87。结论脑结构网络的节点强度差异可以作为一个潜在的影像学生物标志物识别抑郁障碍和双相障碍,联合差异脑区的节点强度可以得到更好的识别率。
