Neural machine translation(NMT)has been widely applied to high-resource language pairs,but its dependence on large-scale data results in poor performance in low-resource scenarios.In this paper,we propose a transfer-l...Neural machine translation(NMT)has been widely applied to high-resource language pairs,but its dependence on large-scale data results in poor performance in low-resource scenarios.In this paper,we propose a transfer-learning-based approach called shared space transfer for zero-resource NMT.Our method leverages a pivot pre-trained language model(PLM)to create a shared representation space,which is used in both auxiliary source→pivot(Ms2p)and(Mp2t)translation models.Specifically,we exploit pivot PLM to initialize the Ms2p decoder pivot→targetand Mp2t encoder,while adopting a freezing strategy during the training process.We further propose a feature converter to mitigate representation space deviations by converting the features from the source encoder into the shared representation space.The converter is trained using the synthetic parallel corpus.The final Ms2t model source→targetcombines the Ms2p encoder,feature converter,and Mp2t decoder.We conduct simulation experiments using English as the pivot language for and translations.We finally test our method German→French,German→Czech,Turkish→Hindion a real zero-resource language pair,with Chinese as the pivot language.Experiment results Mongolian→Vietnameseshow that our method achieves high translation quality,with better Translation Error Rate(TER)and BLEU scores compared with other pivot-based methods.The step-wise pre-training with our feature converter outperforms baseline models in terms of COMET scores.展开更多
设计图像块特征表示是计算机视觉领域内的基本研究内容,优秀的图像块特征表示能够有效地提高图像分类、对象识别等相关算法的性能.SIFT(scale-invariant feature transform)和HOG(histogram of oriented gradient)是人为设计图像块特征...设计图像块特征表示是计算机视觉领域内的基本研究内容,优秀的图像块特征表示能够有效地提高图像分类、对象识别等相关算法的性能.SIFT(scale-invariant feature transform)和HOG(histogram of oriented gradient)是人为设计图像块特征表示的优秀代表,然而,人为设计图像块特征间的差异往往不能足够理想地反映图像块间的相似性.核描述子(kernel descriptor,简称KD)方法提供了一种新的方式生成图像块特征,在图像块间匹配核函数基础上,应用核主成分分析(kernel principal component analysis,简称KPCA)方法进行特征表示,且在图像分类应用上获得不错的性能.但是,该方法需要利用所有联合基向量去生成核描述子特征,导致算法时间复杂度较高.为了解决这个问题,提出了一种算法生成图像块特征表示,称为有效图像块描述子(efficient patch-level descriptor,简称EPLd).算法建立在不完整Cholesky分解基础上,自动选择少量的标志性图像块以提高算法效率,且利用MMD(maximum mean discrepancy)距离计算图像间的相似性.实验结果表明,该算法在图像/场景分类应用中获得了优秀的性能.展开更多
目的:回顾性分析MEDTRONIC Open Pivot主动脉瓣机械瓣置换术后早中期临床结果。方法:连续收集2016年1月至2019年12月在中国医学科学院阜外医院使用MEDTRONIC Open Pivot主动脉瓣机械瓣行瓣膜置换术的患者348例,回顾性分析患者的基线临...目的:回顾性分析MEDTRONIC Open Pivot主动脉瓣机械瓣置换术后早中期临床结果。方法:连续收集2016年1月至2019年12月在中国医学科学院阜外医院使用MEDTRONIC Open Pivot主动脉瓣机械瓣行瓣膜置换术的患者348例,回顾性分析患者的基线临床资料、手术结果、早中期生存率和并发症发生情况。结果:348例患者中,男性283例(81.3%),平均年龄(49.4±11.2)岁。围术期无患者死亡和二次手术。共333例(95.7%)患者完成随访,平均随访(1.9±0.8)年(最长3.36年,共计627人年)。随访期间,11例(3.3%)患者死亡,术后第1、2、3、4年分别死亡3例、2例、5例、1例;5例死于心力衰竭,1例死因为人工瓣膜功能障碍,1例因脑出血死亡,4例为其他非心脑血管原因死亡。术后1年、2年、3年累积生存率分别为99.0%、98.1%、93.5%。随访期间,14例(4.2%)患者出现人工瓣膜功能障碍,发生率为2.23%/人年;6例(1.8%)患者出现瓣周漏,发生率为0.96%/人年;2例(0.6%)患者因严重瓣周漏需二次接受手术。17例(5.1%)患者发生主要不良事件,其中心肌梗死1例(0.3%),脑出血2例(0.6%),脑栓塞1例(0.3%),其他严重出血事件13例(3.9%)。结论:MEDTRONIC Open Pivot主动脉瓣机械瓣有效、耐用,置换术后死亡率及并发症发生率低,是主动脉瓣膜置换的可靠选择之一。展开更多
Cross-spring pivots, formed by crossing two identical flexural beams at their midpoint, have been broadly used in precision engineering and aerospace fields. Many researches have been conducted on modeling and analysi...Cross-spring pivots, formed by crossing two identical flexural beams at their midpoint, have been broadly used in precision engineering and aerospace fields. Many researches have been conducted on modeling and analysis of cross-spring pivots. However the influence of application position and magnitude of the external loads on the load-rotation and parasitic motion characteristics has not yet been discussed. In order to reveal the effect of the external loads, this paper develops the accurate load-rotation and center shift models of cross-spring pivots, with generalized planar loads applied including bending moment, horizontal and vertical forces. Firstly, by using the energy method, the load-displacement models of the pivot are derived with the assumption of small rotational angles. Based on the models, the influence of generalized planar loads on the load-rotation relationship is discussed, which shows that both application position and magnitude of the vertical and horizontal forces influence the load-rotation behaviors. Then the accurate center shift expressions of the pivot with generalized planar loads are developed, which shows that the rotational angle is the dominant term for both components of the center shift while the vertical and horizontal forces are small. Finally, the accuracy of the proposed model is validated by finite element analysis（FEA）. Comparing the model data with the results obtained from FEA, the relative error of the load-rotation is less than 6% even if the rotational angle reaches 20°; the relative errors of the two components of center shift are less than 5% and 10% respectively when the rotational angle reaches 10°. The proposed model and analytical conclusions can be used to analyze and preliminarily design the compliant mechanisms containing cross-spring pivots.展开更多
Large-deflection flexure pivot is widely used in high precision rotation application,but there are less flexure configurations and simple and convenient design methods,This paper presents a novel large-deflection curv...Large-deflection flexure pivot is widely used in high precision rotation application,but there are less flexure configurations and simple and convenient design methods,This paper presents a novel large-deflection curved-compliant annulus-shaped flexure pivot composed of six curved beam flexure elements.It can offer more than lO^angular stroke theoretically.Firstly,main-motion pseudo-rigid-body method is introduced to establish the flexure pivot model.Although pseudo-rigid-body method can be used to analyze the large-deformation flexure pivot performance,the method is definitely a laborious and difficult task for designing this novel flexure pivot.In order to simply the designing process,dimension-design graphs based on the parametric models and finite element analysis is presented.Using the dimension-design method as a tool,the designers can determine the optimal geometry rapidly,based on the stiffness and rotation demands of an annulus-shaped flexure pivot.Finally,dimension-design graph examples are given whose primary design aims to achieve a rotation stroke of annulus-shaped flexure pivot.The finite element analysis results show that the relative designing error between anticipative rotation stroke and graph design result is less than 4%.The dimensionless method used in designing annulus-shaped flexure pivot can reduce design process in both time and complexity.The novel annulus-shaped flexure pivot and dimension-design method are helpful supplement to configuration and design method of large-deflection flexure pivot.展开更多
In concert with developments in global trade and energy resource transportation, there has been a marked increase in reliance on overseas shipping. Unimpeded marine transportation has therefore become a key issue whic...In concert with developments in global trade and energy resource transportation, there has been a marked increase in reliance on overseas shipping. Unimpeded marine transportation has therefore become a key issue which influences national maritime interests including the security of trade and energy resources. A strategic shipping pivot thus performs a vital controlling function for global shipping networks. In this study strategic shipping pivots are defined and subdivided into sea hubs, channels and areas. We then develop a model to identify strategic shipping pivots on a global scale. The results show that, depending on differences in location, function, and type, the concept of strategic shipping pivot permits the identification of both spatial and structural differentiation with respect to strategic hubs, corridors, and seas. Now 44 strategic hubs have formed across the globe. These hubs have become the control centers of local shipping network organization. At the same time, seven strategic corridors containing most shipping routes and transportation capacity connect important sea areas, and permit a high-degree of control over the transport of strategic materials. The strategic seas, the Caribbean, the Mediterranean, Southeast Asia, and the Pacific provide vital import and export pathways, so that the formation of strategic shipping pivots is mainly influenced by factors such as physical geographical conditions, the spatial distribution of socio-economic activities, business organization, technical progress, geopolitical patterns and geopolitical disputes. Physical geographical conditions provide the potential foundations for strategic shipping pivots, while the spatial distribution of socio-economic activities and communications determine the strategic value of these points. Finally, business organization, technical progress, and geopolitical disputes all function to strengthen the strategic mechanisms and the mutagenicity of strategic shipping pivots.展开更多
A direct linear discriminant analysis algorithm based on economic singular value decomposition （DLDA/ESVD） is proposed to address the computationally complex problem of the conventional DLDA algorithm, which directl...A direct linear discriminant analysis algorithm based on economic singular value decomposition （DLDA/ESVD） is proposed to address the computationally complex problem of the conventional DLDA algorithm, which directly uses ESVD to reduce dimension and extract eigenvectors corresponding to nonzero eigenvalues. Then a DLDA algorithm based on column pivoting orthogonal triangular （QR） decomposition and ESVD （DLDA/QR-ESVD） is proposed to improve the performance of the DLDA/ESVD algorithm by processing a high-dimensional low rank matrix, which uses column pivoting QR decomposition to reduce dimension and ESVD to extract eigenvectors corresponding to nonzero eigenvalues. The experimental results on ORL, FERET and YALE face databases show that the proposed two algorithms can achieve almost the same performance and outperform the conventional DLDA algorithm in terms of computational complexity and training time. In addition, the experimental results on random data matrices show that the DLDA/QR-ESVD algorithm achieves better performance than the DLDA/ESVD algorithm by processing high-dimensional low rank matrices.展开更多
基金funded by the National Natural Science Foundation of China(Grant number:Nos.62172341 and 12204386)Sichuan Natural Science Foundation(Grant number:No.2024NSFSC1375)+1 种基金Youth Foundation of Inner Mongolia Natural Science Foundation(Grant number:No.2024QN06017)Basic Scientific Research Business Fee Project for Universities in Inner Mongolia(Grant number:No.0406082215).
文摘Neural machine translation(NMT)has been widely applied to high-resource language pairs,but its dependence on large-scale data results in poor performance in low-resource scenarios.In this paper,we propose a transfer-learning-based approach called shared space transfer for zero-resource NMT.Our method leverages a pivot pre-trained language model(PLM)to create a shared representation space,which is used in both auxiliary source→pivot(Ms2p)and(Mp2t)translation models.Specifically,we exploit pivot PLM to initialize the Ms2p decoder pivot→targetand Mp2t encoder,while adopting a freezing strategy during the training process.We further propose a feature converter to mitigate representation space deviations by converting the features from the source encoder into the shared representation space.The converter is trained using the synthetic parallel corpus.The final Ms2t model source→targetcombines the Ms2p encoder,feature converter,and Mp2t decoder.We conduct simulation experiments using English as the pivot language for and translations.We finally test our method German→French,German→Czech,Turkish→Hindion a real zero-resource language pair,with Chinese as the pivot language.Experiment results Mongolian→Vietnameseshow that our method achieves high translation quality,with better Translation Error Rate(TER)and BLEU scores compared with other pivot-based methods.The step-wise pre-training with our feature converter outperforms baseline models in terms of COMET scores.
文摘设计图像块特征表示是计算机视觉领域内的基本研究内容,优秀的图像块特征表示能够有效地提高图像分类、对象识别等相关算法的性能.SIFT(scale-invariant feature transform)和HOG(histogram of oriented gradient)是人为设计图像块特征表示的优秀代表,然而,人为设计图像块特征间的差异往往不能足够理想地反映图像块间的相似性.核描述子(kernel descriptor,简称KD)方法提供了一种新的方式生成图像块特征,在图像块间匹配核函数基础上,应用核主成分分析(kernel principal component analysis,简称KPCA)方法进行特征表示,且在图像分类应用上获得不错的性能.但是,该方法需要利用所有联合基向量去生成核描述子特征,导致算法时间复杂度较高.为了解决这个问题,提出了一种算法生成图像块特征表示,称为有效图像块描述子(efficient patch-level descriptor,简称EPLd).算法建立在不完整Cholesky分解基础上,自动选择少量的标志性图像块以提高算法效率,且利用MMD(maximum mean discrepancy)距离计算图像间的相似性.实验结果表明,该算法在图像/场景分类应用中获得了优秀的性能.
文摘目的:回顾性分析MEDTRONIC Open Pivot主动脉瓣机械瓣置换术后早中期临床结果。方法:连续收集2016年1月至2019年12月在中国医学科学院阜外医院使用MEDTRONIC Open Pivot主动脉瓣机械瓣行瓣膜置换术的患者348例,回顾性分析患者的基线临床资料、手术结果、早中期生存率和并发症发生情况。结果:348例患者中,男性283例(81.3%),平均年龄(49.4±11.2)岁。围术期无患者死亡和二次手术。共333例(95.7%)患者完成随访,平均随访(1.9±0.8)年(最长3.36年,共计627人年)。随访期间,11例(3.3%)患者死亡,术后第1、2、3、4年分别死亡3例、2例、5例、1例;5例死于心力衰竭,1例死因为人工瓣膜功能障碍,1例因脑出血死亡,4例为其他非心脑血管原因死亡。术后1年、2年、3年累积生存率分别为99.0%、98.1%、93.5%。随访期间,14例(4.2%)患者出现人工瓣膜功能障碍,发生率为2.23%/人年;6例(1.8%)患者出现瓣周漏,发生率为0.96%/人年;2例(0.6%)患者因严重瓣周漏需二次接受手术。17例(5.1%)患者发生主要不良事件,其中心肌梗死1例(0.3%),脑出血2例(0.6%),脑栓塞1例(0.3%),其他严重出血事件13例(3.9%)。结论:MEDTRONIC Open Pivot主动脉瓣机械瓣有效、耐用,置换术后死亡率及并发症发生率低,是主动脉瓣膜置换的可靠选择之一。
基金supported by National Natural Science Foundation of China(Grant Nos. 50975007, 51105014)PhD Programs Foundation of Ministry of Education of China(Grant No. 20091102110023)China Postdoctoral Science Foundation(Grant No. 20100480179)
文摘Cross-spring pivots, formed by crossing two identical flexural beams at their midpoint, have been broadly used in precision engineering and aerospace fields. Many researches have been conducted on modeling and analysis of cross-spring pivots. However the influence of application position and magnitude of the external loads on the load-rotation and parasitic motion characteristics has not yet been discussed. In order to reveal the effect of the external loads, this paper develops the accurate load-rotation and center shift models of cross-spring pivots, with generalized planar loads applied including bending moment, horizontal and vertical forces. Firstly, by using the energy method, the load-displacement models of the pivot are derived with the assumption of small rotational angles. Based on the models, the influence of generalized planar loads on the load-rotation relationship is discussed, which shows that both application position and magnitude of the vertical and horizontal forces influence the load-rotation behaviors. Then the accurate center shift expressions of the pivot with generalized planar loads are developed, which shows that the rotational angle is the dominant term for both components of the center shift while the vertical and horizontal forces are small. Finally, the accuracy of the proposed model is validated by finite element analysis（FEA）. Comparing the model data with the results obtained from FEA, the relative error of the load-rotation is less than 6% even if the rotational angle reaches 20°; the relative errors of the two components of center shift are less than 5% and 10% respectively when the rotational angle reaches 10°. The proposed model and analytical conclusions can be used to analyze and preliminarily design the compliant mechanisms containing cross-spring pivots.
基金supported by National Natural Science Foundation of China(Grant No.50675007)Program for New Century Excellent Talents in University of China(Grant No.NCET-06-0165)
文摘Large-deflection flexure pivot is widely used in high precision rotation application,but there are less flexure configurations and simple and convenient design methods,This paper presents a novel large-deflection curved-compliant annulus-shaped flexure pivot composed of six curved beam flexure elements.It can offer more than lO^angular stroke theoretically.Firstly,main-motion pseudo-rigid-body method is introduced to establish the flexure pivot model.Although pseudo-rigid-body method can be used to analyze the large-deformation flexure pivot performance,the method is definitely a laborious and difficult task for designing this novel flexure pivot.In order to simply the designing process,dimension-design graphs based on the parametric models and finite element analysis is presented.Using the dimension-design method as a tool,the designers can determine the optimal geometry rapidly,based on the stiffness and rotation demands of an annulus-shaped flexure pivot.Finally,dimension-design graph examples are given whose primary design aims to achieve a rotation stroke of annulus-shaped flexure pivot.The finite element analysis results show that the relative designing error between anticipative rotation stroke and graph design result is less than 4%.The dimensionless method used in designing annulus-shaped flexure pivot can reduce design process in both time and complexity.The novel annulus-shaped flexure pivot and dimension-design method are helpful supplement to configuration and design method of large-deflection flexure pivot.
基金The Strategic Priority Research Program of the CAS,No.XDA20010101National Natural Science Foundation of China,No.41571113The Key Project of the Chinese Academy of Sciences,No.ZDRW-ZS-2017-4
文摘In concert with developments in global trade and energy resource transportation, there has been a marked increase in reliance on overseas shipping. Unimpeded marine transportation has therefore become a key issue which influences national maritime interests including the security of trade and energy resources. A strategic shipping pivot thus performs a vital controlling function for global shipping networks. In this study strategic shipping pivots are defined and subdivided into sea hubs, channels and areas. We then develop a model to identify strategic shipping pivots on a global scale. The results show that, depending on differences in location, function, and type, the concept of strategic shipping pivot permits the identification of both spatial and structural differentiation with respect to strategic hubs, corridors, and seas. Now 44 strategic hubs have formed across the globe. These hubs have become the control centers of local shipping network organization. At the same time, seven strategic corridors containing most shipping routes and transportation capacity connect important sea areas, and permit a high-degree of control over the transport of strategic materials. The strategic seas, the Caribbean, the Mediterranean, Southeast Asia, and the Pacific provide vital import and export pathways, so that the formation of strategic shipping pivots is mainly influenced by factors such as physical geographical conditions, the spatial distribution of socio-economic activities, business organization, technical progress, geopolitical patterns and geopolitical disputes. Physical geographical conditions provide the potential foundations for strategic shipping pivots, while the spatial distribution of socio-economic activities and communications determine the strategic value of these points. Finally, business organization, technical progress, and geopolitical disputes all function to strengthen the strategic mechanisms and the mutagenicity of strategic shipping pivots.
基金The National Natural Science Foundation of China (No.61374194)
文摘A direct linear discriminant analysis algorithm based on economic singular value decomposition （DLDA/ESVD） is proposed to address the computationally complex problem of the conventional DLDA algorithm, which directly uses ESVD to reduce dimension and extract eigenvectors corresponding to nonzero eigenvalues. Then a DLDA algorithm based on column pivoting orthogonal triangular （QR） decomposition and ESVD （DLDA/QR-ESVD） is proposed to improve the performance of the DLDA/ESVD algorithm by processing a high-dimensional low rank matrix, which uses column pivoting QR decomposition to reduce dimension and ESVD to extract eigenvectors corresponding to nonzero eigenvalues. The experimental results on ORL, FERET and YALE face databases show that the proposed two algorithms can achieve almost the same performance and outperform the conventional DLDA algorithm in terms of computational complexity and training time. In addition, the experimental results on random data matrices show that the DLDA/QR-ESVD algorithm achieves better performance than the DLDA/ESVD algorithm by processing high-dimensional low rank matrices.
