Previous spectral feature selection methods generate the similarity graph via ignoring the negative effect of noise and redundancy of the original feature space, and ignoring the association between graph matrix learning and feature selection, so that easily producing suboptimal results. To address these issues, this paper joints graph learning and feature selection in a framework to obtain optimal selected performance. More specifically, we use the least square loss function and an ?2,1-norm regularization to remove the effect of noisy and redundancy features, and use the resulting local correlations among the features to dynamically learn a graph matrix from a low-dimensional space of original data. Experimental results on real data sets show that our method outperforms the state-of-the-art feature selection methods for classification tasks.
Multimedia Tools and Applications - Dimensionality reduction techniques based on sparse representation have drawn great attentions recently and they are successfully applied to biometric... 相似文献
Metallurgical and Materials Transactions A - The nucleation potency of iron oxides was verified experimentally through nucleation undercooling of liquid iron using aerodynamic levitation technology... 相似文献
A dual-effect nuclear battery based on the radio-voltaic and radioluminescence effect was developed, which has the ability to convert nuclear energy into electrical energy with two different modes. Performance-enhanced nuclear batteries are mainly based on the addition of ZnS:Cu radio-luminescent layer to Cd-109 X-ray radioactive source and GaAs radio-voltaic layer. In order to explore the response relationship between the mode of energy conversion and the electrical performance of nuclear battery, the physical model was established to research the deposition energy distribution by using Monte Carlo method. The addition of the radio-luminescent material increases the effective energy deposition of the X-rays and the optimized thickness of ZnS:Cu in such a dual-effect nuclear battery should be set to 560 μm. The current–voltage characteristic curves of the batteries before and after performance optimization were utilized to investigate the electrical properties. Through a comprehensive comparison of Cd-109 nuclear batteries with or without radio-luminescent layer, the simulated results are consistent with experimental results. The results indicate that the electrical performance of dual-effect nuclear battery is significantly higher than that of single radio-voltaic nuclear battery. Moreover, the energy conversion efficiency increases from 0.079% (single radio-voltaic nuclear battery) to 0.119% (dual-effect nuclear battery). The improved performance of the dual-effect nuclear battery provides potential applications for space-based autonomous remote sensors and continuous low-power generation technologies. 相似文献
In this study, interface shapes of horizontal oil–water two-phase flow are predicted by using Young-Laplace equation model and minimum energy model. Meanwhile, the interface shapes of horizontal oil–water twophase flow in a 20 mm inner diameter pipe are measured by a novel conductance parallel-wire array probe(CPAP). It is found that, for flow conditions with low water holdup, there is a large deviation between the model-predicted interface shape and the experimentally measured one. Since the variation of pipe wetting characteristics in the process of fluid flow can lead to the changes of the contact angle between the fluid and the pipe wall, the models mentioned above are modified by considering dynamic contact angle. The results indicate that the interface shapes predicted by the modified models present a good consistence with the ones measured by CPAP. 相似文献