Attribute reduction is a research hotspot in rough set theory. Traditional heuristic attribute reduction methods add the most important attribute to the decision attribute set each time, resulting in multiple redundan...Attribute reduction is a research hotspot in rough set theory. Traditional heuristic attribute reduction methods add the most important attribute to the decision attribute set each time, resulting in multiple redundant attribute calculations, high time consumption, and low reduction efficiency. In this paper, based on the idea of sequential three-branch decision classification domain, attributes are treated as objects of three-branch division, and attributes are divided into core attributes, relatively necessary attributes, and unnecessary attributes using attribute importance and thresholds. Core attributes are added to the decision attribute set, unnecessary attributes are rejected from being added, and relatively necessary attributes are repeatedly divided until the reduction result is obtained. Experiments were conducted on 8 groups of UCI datasets, and the results show that, compared to traditional reduction methods, the method proposed in this paper can effectively reduce time consumption while ensuring classification performance.展开更多
Objective High Energy Photon Source-Test Facility(HEPS-TF)is a pre-research project for the construction of high energy synchrotron radiation source in the 12th five-year plan period.The purpose is to research the key...Objective High Energy Photon Source-Test Facility(HEPS-TF)is a pre-research project for the construction of high energy synchrotron radiation source in the 12th five-year plan period.The purpose is to research the key technology and develop the key equipment of high energy synchrotron radiation source.Superconducting 3W1 magnet is the first self-developed superconducting wiggler magnet in China,and it is also one of the key research topics of HEPS.The author has completed a new digital closed-loop control algorithm for the superconducting 3W1 magnet with large load time constant and the nonlinear characteristics of inductance increasing with current,namely three-branch structure algorithm.In the face of the rapid development of high energy accelerator technology,the application of intelligent technology has become an inevitable development trend in the field of accelerator magnet power supply technology.Although the digital control of accelerator magnet power supply has been widely used,it is the first time to apply the new closed-loop control algorithm to realize fast adjustment and precision tracking in accelerator superconducting magnet power supply in China.Method According to the nonlinear characteristics of inductance and output current of superconducting magnet,a new digital closed-loop control algorithm for the load of superconducting magnet power supply with large time constant is proposed.Conclusion This algorithm is quite different from the traditional algorithm and can attain the independent tracking and adjust-ment of the control target.Finally,by testing the ripple,error and stability of superconducting 3W1 magnet power supply,the correctness,practicability and reliability of power supply system as well as the digital control algorithm are verified.The results provide a new idea for the control of accelerator magnet power supply.展开更多
针对传统目测法检测贴片二极管表面缺陷效率低下和基于手工特征的目标检测算法模型较浅,以及语义性不高等问题,提出了改进YOLO-V4的贴片二极管表面缺陷检测方法。首先考虑到随着网络加深使梯度消失,以及减少网络中的特征冗余和参数量的...针对传统目测法检测贴片二极管表面缺陷效率低下和基于手工特征的目标检测算法模型较浅,以及语义性不高等问题,提出了改进YOLO-V4的贴片二极管表面缺陷检测方法。首先考虑到随着网络加深使梯度消失,以及减少网络中的特征冗余和参数量的情况,CSP1模块采用DenseNet替换原网络中的ResNet;其次,为了实现特征信息的跨维度交互,让网络更加关注重要信息,在CSP1模块后引入了三分支注意力机制模块,同时使用FPN+PANet对特征进行融合;并且用CSP2替换CBL×5模块,降低了网络的运算量,提高了算法检测速度;最后优化了Focal Loss函数,对正负样本添加权重,以解决正负样本不平衡的问题。本文算法相较于YOLO-V4的检测精度(precision,P)、召回率(recall,R)和多分类平均精度(mean average precision,mAP),分别高出2.98%,2.65%,2.92%,表明改进YOLO-V4可以有效检测贴片二极管表面缺陷问题。展开更多
文摘Attribute reduction is a research hotspot in rough set theory. Traditional heuristic attribute reduction methods add the most important attribute to the decision attribute set each time, resulting in multiple redundant attribute calculations, high time consumption, and low reduction efficiency. In this paper, based on the idea of sequential three-branch decision classification domain, attributes are treated as objects of three-branch division, and attributes are divided into core attributes, relatively necessary attributes, and unnecessary attributes using attribute importance and thresholds. Core attributes are added to the decision attribute set, unnecessary attributes are rejected from being added, and relatively necessary attributes are repeatedly divided until the reduction result is obtained. Experiments were conducted on 8 groups of UCI datasets, and the results show that, compared to traditional reduction methods, the method proposed in this paper can effectively reduce time consumption while ensuring classification performance.
基金the Accelerator Centre of Insti-tute of High Energy Physics for financial support
文摘Objective High Energy Photon Source-Test Facility(HEPS-TF)is a pre-research project for the construction of high energy synchrotron radiation source in the 12th five-year plan period.The purpose is to research the key technology and develop the key equipment of high energy synchrotron radiation source.Superconducting 3W1 magnet is the first self-developed superconducting wiggler magnet in China,and it is also one of the key research topics of HEPS.The author has completed a new digital closed-loop control algorithm for the superconducting 3W1 magnet with large load time constant and the nonlinear characteristics of inductance increasing with current,namely three-branch structure algorithm.In the face of the rapid development of high energy accelerator technology,the application of intelligent technology has become an inevitable development trend in the field of accelerator magnet power supply technology.Although the digital control of accelerator magnet power supply has been widely used,it is the first time to apply the new closed-loop control algorithm to realize fast adjustment and precision tracking in accelerator superconducting magnet power supply in China.Method According to the nonlinear characteristics of inductance and output current of superconducting magnet,a new digital closed-loop control algorithm for the load of superconducting magnet power supply with large time constant is proposed.Conclusion This algorithm is quite different from the traditional algorithm and can attain the independent tracking and adjust-ment of the control target.Finally,by testing the ripple,error and stability of superconducting 3W1 magnet power supply,the correctness,practicability and reliability of power supply system as well as the digital control algorithm are verified.The results provide a new idea for the control of accelerator magnet power supply.
文摘针对传统目测法检测贴片二极管表面缺陷效率低下和基于手工特征的目标检测算法模型较浅,以及语义性不高等问题,提出了改进YOLO-V4的贴片二极管表面缺陷检测方法。首先考虑到随着网络加深使梯度消失,以及减少网络中的特征冗余和参数量的情况,CSP1模块采用DenseNet替换原网络中的ResNet;其次,为了实现特征信息的跨维度交互,让网络更加关注重要信息,在CSP1模块后引入了三分支注意力机制模块,同时使用FPN+PANet对特征进行融合;并且用CSP2替换CBL×5模块,降低了网络的运算量,提高了算法检测速度;最后优化了Focal Loss函数,对正负样本添加权重,以解决正负样本不平衡的问题。本文算法相较于YOLO-V4的检测精度(precision,P)、召回率(recall,R)和多分类平均精度(mean average precision,mAP),分别高出2.98%,2.65%,2.92%,表明改进YOLO-V4可以有效检测贴片二极管表面缺陷问题。
