Time filter

Source Type

Yin M.,Guangdong University of Technology | Yin M.,Key Laboratory of Autonomous Systems and Networked Control | Gao J.,Charles Sturt University | Lin Z.,Peking University | Lin Z.,Shanghai JiaoTong University
IEEE Transactions on Pattern Analysis and Machine Intelligence | Year: 2016

Low-rank representation (LRR) has recently attracted a great deal of attention due to its pleasing efficacy in exploring low-dimensional subspace structures embedded in data. For a given set of observed data corrupted with sparse errors, LRR aims at learning a lowest-rank representation of all data jointly. LRR has broad applications in pattern recognition, computer vision and signal processing. In the real world, data often reside on low-dimensional manifolds embedded in a high-dimensional ambient space. However, the LRR method does not take into account the non-linear geometric structures within data, thus the locality and similarity information among data may be missing in the learning process. To improve LRR in this regard, we propose a general Laplacian regularized low-rank representation framework for data representation where a hypergraph Laplacian regularizer can be readily introduced into, i.e., a Non-negative Sparse Hyper-Laplacian regularized LRR model (NSHLRR). By taking advantage of the graph regularizer, our proposed method not only can represent the global low-dimensional structures, but also capture the intrinsic non-linear geometric information in data. The extensive experimental results on image clustering, semi-supervised image classification and dimensionality reduction tasks demonstrate the effectiveness of the proposed method. © 1979-2012 IEEE. Source

Zhang Y.,Sun Yat Sen University | Zhai K.,Sun Yat Sen University | Chen D.,SYSU CMU Shunde International Joint Research Institute | Jin L.,SYSU CMU Shunde International Joint Research Institute | Hu C.,Key Laboratory of Autonomous Systems and Networked Control
Mathematics and Computers in Simulation | Year: 2015

Zhang-gradient (ZG) method is a combination of Zhang dynamics (ZD) and gradient dynamics (GD) methods which are two powerful methods for online time-varying problems solving. ZG controllers are designed using the ZG method to solve the tracking control problem of a certain system. In this paper, the design process of the ZG controllers with explicit as well as implicit tracking control of the double-integrator system is presented in detail. In addition, the corresponding computer simulations are conducted with different values of the design parameter λ to illustrate the effectiveness of ZG controllers. However, even though the ZG controllers are powerful, there is still a challenge in the simulation practice. Specifically, different settings of simulation options in MATLAB ordinary differential equation (ODE) solvers may lead to different simulation results (e.g.,failure and success). For better comparison, the successful and failed simulation results are both presented. The differences in simulation results remind us to pay more attention to MATLAB defaults and options when we conduct such simulations. © 2015 International Association for Mathematics and Computers in Simulation (IMACS). Source

Zhang Y.,Sun Yat Sen University | Zhang Y.,SYSU CMU Shunde International Joint Research Institute | Zhang Y.,Key Laboratory of Autonomous Systems and Networked Control | Qu L.,Sun Yat Sen University | And 5 more authors.
Soft Computing | Year: 2016

To solve complex problems such as multi-input function approximation by using neural networks and to overcome the inherent defects of traditional back-propagation neural networks, a single hidden-layer feed-forward sine-activated neural network, sine neural network (SNN), is proposed and investigated in this paper. Then, a double-stage weights and structure determination (DS-WASD) method, which is based on the weights direct determination method and the approximation theory of using linearly independent functions, is developed to train the proposed SNN. Such a DS-WASD method can efficiently and automatically obtain the relatively optimal SNN structure. Numerical results illustrate the validity and efficacy of the SNN model and the DS-WASD method. That is, the proposed SNN model equipped with the DS-WASD method has great performance of approximation on multi-input function data. © 2014, Springer-Verlag Berlin Heidelberg. Source

Qiu B.,Sun Yat Sen University | Qiu B.,SYSU CMU Shunde International Joint Research Institute | Qiu B.,Key Laboratory of Autonomous Systems and Networked Control | Zhang Y.,Sun Yat Sen University | And 4 more authors.
Advanced Robotics | Year: 2016

In this paper, by revisiting Ma et al.’s inspiring work (specifically, Ma equivalence, ME) and Zhang et al.’s inspiring work (specifically, Zhang equivalence, ZE), which both investigate the equivalence relationships of redundancy-resolution schemes at two different levels, but with different formulations, the general scheme formulations and equivalence analyses of ME and ZE are presented. Besides, being a case study, the ME and ZE of minimum velocity norm (MVN) type are investigated for the inverse-kinematics (IK) problem solving. Moreover, the link and difference between the MVN-type ME and ZE are analyzed, summarized and presented methodologically, systematically, and computationally in this paper. In order to numerically compare the ME and ZE of MVN type, a Rhodonea-path tracking task based on PUMA560 robot manipulator is tested and fulfilled by employing the original velocity-level MVN schemes and its equivalent acceleration-level MVN schemes of ME and ZE. The simulative and numerical results not only verify the effectiveness of the velocity-level and acceleration-level schemes of MVN-type ME and ZE, but also validate the reasonableness of such two proved equivalence relationships. More importantly, these results show quantitatively and comparatively the respective advantages and future applications of MVN-type ME and ZE for the IK problem solving. © 2016 Taylor & Francis and The Robotics Society of Japan Source

Jin L.,Sun Yat Sen University | Jin L.,SYSU CMU Shunde International Joint Research Institute | Jin L.,Key Laboratory of Autonomous Systems and Networked Control | Zhang Y.,Sun Yat Sen University | And 6 more authors.
Neurocomputing | Year: 2016

The tracking-control problem of a special nonlinear system (i.e., the extension of a modified Lorenz chaotic system) with additive input or the mixture of additive and multiplicative inputs is considered in this paper. It is worth pointing out that, with the parameters fixed at some particular values, the modified Lorenz nonlinear system degrades to the modified Lorenz chaotic system. Note that, due to the existence of singularities at which the nonlinear system fails to have a well-defined relative degree, the input-output linearization method and the backstepping design technique cannot solve the tracking-control problem. By combining Zhang neural dynamics and gradient neural dynamics, a new effective controller-design method, termed Zhang-gradient (ZG) neural dynamics, is proposed for the tracking control of the modified Lorenz nonlinear system. With singularities conquered, this ZG neural dynamics is able to solve the tracking-control problem of the modified Lorenz nonlinear system via additive input or mixed inputs (i.e., the mixture of additive and multiplicative inputs). Both theoretical analyses and simulative verifications substantiate that the tracking controllers based on the ZG neural dynamics with additive input or mixed inputs not only achieve satisfactory tracking accuracy but also successfully conquer the singularities encountered during the tracking-control process. Moreover, the applications to the synchronization, stabilization and tracking control of other nonlinear systems further illustrate the effectiveness and advantages of the ZG neural dynamics. © 2016 Elsevier B.V. Source

Discover hidden collaborations