1,256 publications from this institution
Hamid Karimi, Jiliang Tang. Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers). 2019.
No abstract is provided for this article.
Due to its simplicity and easy implementation, partial least squares (PLS) serves as an efficient approach in large-scale industrial process. However, like many data-based methods, PLS is quite sensitive to outliers, which is a common abnormal characteristic of the measured process data that can significantly affect the monitoring performance of PLS. In order to develop a robust prediction and fault detection method, this paper employs the partial robust M-regression (PRM) to deal with the outliers. Moreover, to eliminate the useless variations for prediction, an orthogonal decomposition is performed on the measurable variables space so as to allow the new method to serve as a powerful tool for quality-related prediction and fault detection. The proposed method is finally applied on the Tennessee Eastman (TE) process.
Driving behavior is crucial to the energy consumption analysis of electric vehicles. This paper proposes an unsupervised learning method to classify driving behavior for three typical road conditions. First, three specific road conditions are selected from the open access data, including characteristic information such as speed and acceleration. Besides, the characteristic data is processed, so each distinct value has the same weight. Second, two unsupervised learning clustering algorithms are introduced and compared in typical working conditions. Finally, the clustering results under three working conditions are obtained. Specifically, we can classify driving styles in high-speed conditions into aggressive, standard, and calm; besides, the classification method of K-medoids is more advantageous. In intersection conditions, driving styles are usually divided into standard and calm. Considering the calculation time and other factors, the K-means algorithm shows superior effects compared to the K-medoids algorithm. The driving style can be divided into standard and calm in campus conditions. In this case, K-medoids have a more significant advantage. The research results have implications for the classification of driving styles under different road conditions.
International audience
The paper is concerned with the application of wavelet-based neural networks for optimal control of robotic manipulators motion. Optimal control law is found by optimization of Hamilton-Jacobi-Bellman (H-J-B) equation and it is shown how wavelet-based neural networks can overcome nonlinearities through optimization with no preliminary off-line learning phase required. The neural network is learned as on-line and the adaptive learning algorithm is derived from Lyapunov stability analysis. So that both system tracking stability and error convergence of nonlinear function estimating can be guaranteed in the closed-loop system. The Lyapunov function for the nonlinear analysis is derived from the user input in terms of a specified quadratic performance index. Simulation results illustrate the effectiveness of our method.
In this paper we present an implementation of the Haar wavelet to the optimal control of linear singularly perturbed systems. The approximated composite control and the slow and fast trajectories with respect to a quadratic cost function are calculated by solving only the linear algebraic equations. The results are illustrated with a simple example.
No abstract is provided for this article.
No abstract is provided for this article.
Codebook is an effective image representation method. By clustering in local image descriptors, a codebook is shown to be a distinctive image feature and widely applied in object classification. In almost all existing works on codebooks, the building of the visual vocabulary follows a basic routine, that is, extracting local image descriptors and clustering with a user-designated number of clusters. The problem with this routine lies in that building a codebook for each single dataset is not efficient. In order to deal with this problem, we investigate the influence of vocabulary sizes on classification performance and vocabulary universality with the kNN classifier. Experimental results indicate that, under the condition that the vocabulary size is large enough, the vocabularies built from different datasets are exchangeable and universal.
Notice of Violation of IEEE Publication Principles <br><br>After careful consideration by a duly constituted committee, an author of this article, Hamid Reza Karimi, was found to have acted in violation of the IEEE Principles of Ethical Publishing by artificially inflating the number of citations to this article. <br/> This article studies the robust and reliable H <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">∞</sub> static output feedback (SOF) control for nonlinear systems with actuator faults in a descriptor system framework. The nonlinear plant is characterized by a discrete-time Takagi-Sugeno (T-S) fuzzy affine model with parameter uncertainties, and the Markov chain is utilized to describe the actuator-fault behaviors. Specifically, by adopting a state-output augmentation approach, the original system is firstly reformulated into the descriptor fuzzy affine system. Based upon a novel piecewise Markovian Lyapunov function (LF), the H <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">∞</sub> performance analysis condition for the underlying system is then presented, and furthermore the robust and reliable SOF controller synthesis is carried out. It is shown that by invoking the redundancy properties induced by the descriptor formulation, combined with some convexifying techniques, the existence of the desired reliable controller can be explicitly determined by the solution of a convex optimization problem. Finally, simulation studies are applied to confirm the effectiveness of the developed method.
In this paper, we present a novel control design methodology for structural vibration control of large buildings. The main idea consists in decomposing the overall building system into decoupled single-story subsystems and modeling the subsystem interactions as external disturbances. Then, a complete set of local decentralized controllers can be efficiently computed using the existing LMI solvers. In the proposed approach, two different levels of decentralization can be distinguished: decentralized design and decentralized implementation, which are both of critical importance in large-scale control problems. From the design point of view, the local controllers are independently synthesized using only the low-dimensional subsystem models. The implementation phase is through the overall decentralized controller defined by the set of local controllers, which can drive the actuation devices using only local state-feedback information. To illustrate the proposed methodology, decentralized H8 controllers are designed for the seismic protection of a five-story building and a twenty-story building. A proper set of numerical simulations is carried out to demonstrate the effectiveness of the proposed decentralized controllers and the computation times are considered to assess the computational effectiveness of the decentralized design methodology.
Feature combination is a powerful approach to improve object classification performance. While various combination algorithms have been proposed, average combination is almost always selected as the baseline algorithm to be compared with. In previous work we have found that it is better to use only a sample of the most powerful features in average combination than using all. In this paper, we continue this work and further show that the behaviors of features in average combination can be integrated into the k -Nearest-Neighbor (kNN) framework. Based on the kNN framework, we then propose to use a selection based average combination algorithm to obtain the best classification performance from average combination. Our experiments on four diverse datasets indicate that this selection based average combination performs evidently better than the ordinary average combination, and thus serves as a better baseline. Comparing with this new and better baseline makes the claimed superiority of newly proposed combination algorithms more convincing. Furthermore, the kNN framework is helpful in understanding the underlying mechanism of feature combination and motivating novel feature combination algorithms.
