中国科学院机构知识库网格系统: 检索

Accelerating Minibatch Stochastic Gradient Descent Using Typicality Sampling 期刊论文 OAI收割
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS, 2020, 卷号: 31, 期号: 11, 页码: 4649-4659
作者: Peng, Xinyu; Li, Li; Wang, Fei-Yue

|

收藏 | 浏览/下载：21/0 | 提交时间：2021/01/06

Training Convergence Approximation algorithms Stochastic processes Estimation Optimization Acceleration Batch selection machine learning minibatch stochastic gradient descent (SGD) speed of convergence

A new algorithm for solving the best-fit sphere of optical aspherical surface (EI CONFERENCE) 会议论文 OAI收割
2nd International Conference on Advances in Materials and Manufacturing, ICAMMP 2011, December 16, 2011 - December 18, 2011, Guilin, China
作者: Lin J.; Lu M.

收藏 | 浏览/下载：40/0 | 提交时间：2013/03/25

To solve the best-fit sphere (BFS) accurately is one of the technological keys for the generating and testing of optical aspherical surfaces. This paper presents a new algorithm for solving the BFS of aspherical surfaces to suppress some deficiencies in the existing BFS algorithms. In the proposed approach it is not only suitable for the conic surface a BFS is constructed but also for higher order aspheres. The obtained asphericity and material removal function is more suitable for the machining and test. (2012) Trans Tech Publications which passes through both sides of endpoints in the section o -?xy of the aspherical surfaces Switzerland. the center of the BFS is shifted along the x-axis and its radius of curvature is automatically computed. The variable step size method is proposed to speed up the convergence of the iteration. Through numerically solving the BFS of conic and cubic surface the advantages of the proposed approach are verified. The results show that the proposed approach is of rapid convergence and high accuracy

Research on satellite platform vibration compensation method based on liquid crystal steering system (EI CONFERENCE) 会议论文 OAI收割
2011 SSITE International Conference on Future Material Research and Industry Application, FMRIA 2011, December 1, 2011 - December 2, 2011, Macau, China
作者: Wen M.

收藏 | 浏览/下载：16/0 | 提交时间：2013/03/25

This paper presents the disturbance compensation scheme based on liquid crystal steering system according to the problem that satellite platform vibration affects the tracking and pointing accuracy in space optical communication. The mathematical model of liquid crystal steering system is built. For improving convergence speed the adaptive controller based on posteriori estimation errors RLS lattice filter is designed which not only achieves fast adaptation but also reaches optimal steady-state disturbance compensation. Finally the simulation results show that this control scheme is valid and feasible. (2012) Trans Tech Publications Switzerland.

Research on satellite platform vibration compensation technology for space optical communication (EI CONFERENCE) 会议论文 OAI收割
2012 International Meeting on Opto-Electronics Engineering and Materials Research, OEMR 2012, July 27, 2012 - July 29, 2012, Shenyang, Liaoning, China
Han C.; Li X.

收藏 | 浏览/下载：22/0 | 提交时间：2013/03/25

The piezo-electric steering control system is used to compensate the vibration of satellite platform. The mathematical model of piezo-electric steering system is built. The adaptive controller based on improved LMS filter algorithm is designed according to this system characteristic. The adaptive delay filter is designed to solve slow convergence speed problem of traditional algorithm. Meanwhile the adaptive delay filter can also compensate phase delay of control circuit. The adaptive deviation filter is introduced and it effectively compensates the vibration deviation which is low frequency and high amplitude. The experimental results show that the remained errors are rad which meets space optical communication requirement and proves that this control scheme is valid and feasible. (2012) Trans Tech Publications Switzerland.

The application of symmetrical equation of diffracted wavevectors in integral theory of gratings (EI CONFERENCE) 会议论文 OAI收割
2nd International Conference on Advanced Engineering Materials and Technology, AEMT 2012, July 6, 2012 - July 8, 2012, Zhuhai, China
作者: Gao J.; Zhang F.; Zhang F.

收藏 | 浏览/下载：17/0 | 提交时间：2013/03/25

The integral theory for electromagnetic theory of gratings which has complex mathematical tool and long computation cost is needed to improve computing speed of the algorithm. A very shot-cut equation which is called symmetrical equation of diffracted wavevectors by us is presented. It makes the number of Green's function expansion terms shorter and simplifies the integral kernel by taking into account its symmetry. Take the grating in Littrow mount for computing example because of applying this equation the computation cost of integral theory which has good convergence is shorten to 1/2 to 2/5. This work modifies the integral theory and improves its computing speed. (2012) Trans Tech Publications Switzerland.

Image matching using a bat algorithm with mutation (EI CONFERENCE) 会议论文 OAI收割
2012 International Conference on Mechatronic Systems and Automation Systems, MSAS 2012, July 21, 2012 - July 21, 2012, Wuhan, China
作者:

收藏 | 浏览/下载：39/0 | 提交时间：2013/03/25

Due to shortcoming of traditional image matching for computing the fitness for every pixel in the searching space a new bat algorithm with mutation (BAM) is proposed to solve image matching problem and a modification is applied to mutate between bats during the process of the new solutions updating. This new approach can accelerate the global convergence speed while preserving the strong robustness of the basic BA. The realization procedure for this improved meta-heuristic approach BAM is also presented. To prove the performance of this proposed meta-heuristic method BAM is compared with BA and other population-based optimization methods DE and SGA. The experiment shows that the proposed approach is more effective and feasible in image matching than the other model. (2012) Trans Tech Publications Switzerland.

Application of improved UKF algorithm in initial alignment of SINS (EI CONFERENCE) 会议论文 OAI收割
2011 2nd International Conference on Artificial Intelligence, Management Science and Electronic Commerce, AIMSEC 2011, August 8, 2011 - August 10, 2011, Zhengzhou, China
Su W. X.

收藏 | 浏览/下载：20/0 | 提交时间：2013/03/25

In order to improve the initial alignment accuracy and convergence rate of the SINS system proposed the improved UKF algorithm (AUKF) based on the Unscented Kalman Filter (UKF). Noise statistical characteristics are mostly unknown in real systems when it was effected by the initial value errors and dynamic model errors AUKF algorithm can real-time adjust the covariance of the state vector and observation vector and balance the right ratio of the state information and observation information in the filter results thereby improving the system performance. The experimental results show: The Improved UKF Algorithm enhances the convergence speed and alignment accuracy effectively. 2011 IEEE.

Research on the identification for a nonlinear system (EI CONFERENCE) 会议论文 OAI收割
International Conference on Optical, Electronic Materials and Applications 2011, OEMA 2011, March 4, 2011 - March 6, 2011, Chongqing, China
作者: Liu J.; Jia P.; Liu J.; Liu J.

收藏 | 浏览/下载：32/0 | 提交时间：2013/03/25

The characteristic of the drift error of inertial platform is a high-order nonlinear dynamic system using the neural networks' abilities of universal approximation of differentiable trajectory and capturing system dynamic information this paper presents the drift error identifying project of inertial platform based on Elman networks structure. First the drift error model of inertial platform is established after selecting the input and output for network momentum and alterable speed algorithm is used to speed up the network convergence. On the basis of the algorithm the extended nonlinear node function in the hidden network does not only improve the learning speed of network but also satisfies the need of accuracy on system identification. Through the drift error data measured on inertial platform the training result shows that the scheme achieves satisfied identification results. (2011) Trans Tech Publications.

Application of adaptive Kalman filter technique in initial alignment of strapdown inertial navigation system (EI CONFERENCE) 会议论文 OAI收割
29th Chinese Control Conference, CCC'10, July 29, 2010 - July 31, 2010, Beijing, China
作者: Liu P.

收藏 | 浏览/下载：28/0 | 提交时间：2013/03/25

In order to improve the alignment precision and convergence speed of strap-down inertial navigation system but in the active system most noise statistical characteristics are unknown an initial alignment method based on Sage-Husa adaptive filter is presented. We also derived the exactitude alignment error model and adaptive Kalman filter equation in the azimuth of small misalignment angle. As usual in this case known the noise statistical characteristics we introduce the adaptive Kalman filter. It uses the information of observed data Kalman filter is suitable on-line estimation noise statistical characteristics and state simultaneously in order to improve the filter continuously so the filter has a higher estimation accuracy than the conventional Kalman filter. By simulating verifying the adaptive Kalman filter enhances the convergence speed and alignment accuracy effectively.

The registration of aerial infrared and visible images (EI CONFERENCE) 会议论文 OAI收割
2010 International Conference on Educational and Information Technology, ICEIT 2010, September 17, 2010 - September 19, 2010, Chongqing, China
作者: Liu J.; Wang Y.; Wang Y.; Wang Y.; Wang Y.

收藏 | 浏览/下载：27/0 | 提交时间：2013/03/25

In order to solve the registration problem of different source image existed on aerial image fusion algorithms based on Particle Swarm Optimization (PSO) are applied as search strategy in this paper and Alignment Metric (AM) is used as judgment. This study has realized the different source image registration of infrared and visible light with high speed high accuracy and high reliability. Basically with little restriction of gray level properties a new alignment measure is applied which can efficiently measure the image registration extent and tolerate noise well. Even more the intelligent optimization algorithm - Particle Swarm Optimization (PSO) is combined to improve the registration precision and rate of infrared and visible light. Experimental results indicate that the study attains the registration accuracy of pixel level and every registration time is cut down over 40 percent compared to traditional method. The match algorithm based on AM solves the registration problem that greater differences between different source images are existed on gray and characteristic. At the same time the adoption of combining the intelligent optimization algorithms significantly improves the searching efficiency and convergence speed of the algorithms and the registration result has higher accuracy and stability which builds up solid foundation for different source image fusion. The method in this paper has a magnificent effect and is easy for application and very suitable for engineering use. 2010 IEEE.