Locality Preserving Maximum Scatter Difference Projection for Face Recognition

Maximum Scatter Difference (MSD) aims to preserve discriminant information of sample space, but it fails to find the essential structure of the samples with nonlinear distribution. To overcome this problem, an efficient feature extraction method named as Locality Preserving Maximum Scatter Difference (LPMSD) projection is proposed in this paper. The new algorithm is developed based on locality preserved embedding and MSD criterion. Thus, the proposed LPMSD not only preserves discriminant information of sample space but also captures the intrinsic submanifold of sample space. Experimental results on ORL, Yale and CAS-PEAL face database indicate that the LPMSD method outperforms the MSD, MMSD and LDA methods under various experimental conditions.

Observed Tropospheric Temperature Response to 11-yr Solar Cycle and What It Reveals about Mechanisms

Using 54 yr of NCEP reanalysis global data from 1000 to 10 hPa, this study establishes the existence and the statistical significance of the zonal-mean temperature response to the 11-yr solar cycle throughout the troposphere and parts of the lower stratosphere. Two types of statistical analysis are used: the composite-mean difference projection method, which tests the existence of the solar cycle signal level by level, and the adaptive AR(p)-t test, which tells if a particular local feature is statistically significant at the 95% confidence level. A larger area of statistical significance than that in previous published work is obtained, due to the longer record and a better trend removal process. It reveals a spatial pattern consistent with a “bottom up” mechanism, involving evaporative feedback near the tropical ocean surface and tropical vertical convection, latent heating of the tropical upper troposphere, and poleward large-scale heat transport to the polar regions. It provides an alternative to the currently favored “top down” mechanism involving stratospheric ozone heating.