Papers by Keyword: Manifold Learning

Abstract: Locally linear embedding is based on the assumption that the whole data manifolds are evenly distributed so that they determine the neighborhood for all points with the same neighborhood size. Accordingly, they fail to nicely deal with most real problems that are unevenly distributed. This paper presents a new approach that takes the general conceptual framework of Hessian locally linear embedding so as to guarantee its correctness in the setting of local isometry to an open connected subset but dynamically determines the local neighborhood size for each point. This approach estimates the approximate geodesic distance between any two points by the shortest path in the local neighborhood graph, and then determines the neighborhood size for each point by using the relationship between its local estimated geodesic distance matrix and local Euclidean distance matrix. This approach has clear geometry intuition as well as the better performance and stability to deal with the sparsely sampled or noise contaminated data sets that are often unevenly distributed. The conducted experiments on benchmark data sets validate the proposed approach.

Abstract: Ranking data points with respect to a given preference criterion is an example of a preference learning task. In this paper, we investigate the generalization performance of the regularized ranking algorithm associated with least square ranking loss in a reproducing kernel Hilbert space, and use the method of computing hold-out estimates for the proposed algorithm. Based on using the hold-out method, we obtain fast learning rate for this algorithm.

Abstract: Fault diagnosis for wind turbine is an important task for reducing their maintenance cost. However, the non-stationary dynamic operating conditions of wind turbines pose a challenge to fault diagnosis for wind turbine. Fault diagnosis is essentially a kind of pattern recognition. In this paper, a novel fault diagnosis method based on enhanced supervised locally linear embedding is proposed for wind turbine. The approach first performs the recently proposed manifold learning algorithm locally linear embedding on the high-dimensional fault signal samples to learn the intrinsic embedded multiple manifold features corresponding to different fault modes. Enhanced supervised locally linear embedding not only can map them into a low-dimensional embedded space to achieve fault feature extraction, but also can deal with new fault samples. Finally fault classification is carried out in the embedded manifold space. The wind turbine gearbox ball bearing vibration fault signals are used to validate the proposed fault diagnosis method. The results indicate that the proposed approach obviously improves the fault classification performance and outperforms the other traditional approaches.

Abstract: A face recognition method on mobile terminals based on manifold learning was proposed. Firstly, the modified Snake model was set in order to improve the accuracy and effectiveness of facial feature point labeling. Then, the partial mapping method was carried out to map the face images to a subspace for further analysis. Finally, the nearest neighbor classifier was enhanced to show the recognition results. The experimental results indicate that the performance of this method is excellent. It is boasts a higher accuracy rate and bigger robustness than the ordinary methods.

Abstract: To improve effectively the performance on speech emotion recognition, it is needed to perform nonlinear dimensionality reduction for speech feature data lying on a nonlinear manifold embedded in high-dimensional acoustic space. This paper proposes an improved SLLE algorithm, which enhances the discriminating power of low-dimensional embedded data and possesses the optimal generalization ability. The proposed algorithm is used to conduct nonlinear dimensionality reduction for 48-dimensional speech emotional feature data including prosody so as to recognize three emotions including anger, joy and neutral. Experimental results on the natural speech emotional database demonstrate that the proposed algorithm obtains the highest accuracy of 90.97% with only less 9 embedded features, making 11.64% improvement over SLLE algorithm.

Abstract: Self-adapt distance measure supervised locally linear embedding solves the problem that Euclidean distance measure can not apart from samples in content-based image retrieval. This method uses discriminative distance measure to construct k-NN and effectively keeps its topological structure in high dimension space, meanwhile it broadens interval of samples and strengthens the ability of classifying. Experiment results show the ADM-SLLE date-reducing-dimension method speeds up the image retrieval and acquires high accurate rate in retrieval.

Abstract: Manifold learning has made many successful applications in the fields of dimensionality reduction, pattern recognition, and data visualization. In this paper we proposed hierarchical macro manifold (HMM) for the purpose of supervised classification. We construct hierarchical macro manifold based on the given training sets. The generalized regression neural network is employed to solve the out-of-sample problem. Experimental results demonstrate the feasibility and effectiveness of our proposed approach.

Abstract: In this paper, we propose the motion sequence segmentation based on LLE (Locally Linear Embedding) algorithm. The method is to reduce the dimension of the high dimension motion sequence to obtain one-dimension feature curve. Then we use the feature curve to achieve motion sequence segmentation. Simulation results demonstrate that this method can achieve motion sequences segmentation and improve the accuracy rate greatly compared with the traditional algorithm.

Abstract: A motion key-frames extraction algorithm based on LLE was presented. First, the dimensionality of motion capture data was reduced by LLE method to obtain the low dimension feature curve. Second, the initial key-frames were obtained by extracting the local extreme points. In the end, according to the difference of feature curve amplitude and the threshold, the corresponding key-frames were inserted to get the final key-frames. A amount of experiments showed that the algorithm could generalize the content of original motion sequence on the basis of the compression ratio and low error rate.

Abstract: Fault diagnosis is essentially a kind of pattern recognition. In this paper propose a novel machinery fault diagnosis method based on supervised locally linear embedding is proposed first. The approach first performs the recently proposed manifold learning algorithm locally linear embedding on the high-dimensional fault signal samples to learn the intrinsic embedded multiple manifold features corresponding to different fault modes. Supervised locally linear embedding not only can map them into a low-dimensional embedded space to achieve fault feature extraction, but also can deal with new fault samples. Finally fault classification is carried out in the embedded manifold space. The ball bearing fault signals are used to validate the proposed fault diagnosis method. The results indicate that the proposed approach obviously improves the fault classification performance and outperforms the other traditional approaches.