Sparse kernel logistic regression based on L 1/2 regularization

The sparsity driven classification technologies have attracted much attention in recent years, due to their capability of providing more compressive representations and clear interpretation. Two most popular classification approaches are support vector machines (SVMs) and kernel logistic regression (KLR), each having its own advantages. The sparsification of SVM has been well studied, and many sparse versions of 2-norm SVM, such as 1-norm SVM (1-SVM), have been developed. But, the sparsification of KLR has been less studied. The existing sparsification of KLR is mainly based on L ₁ norm and L ₂ norm penalties, which leads to the sparse versions that yield solutions not so sparse as it should be. A very recent study on L _1/2 regularization theory in compressive sensing shows that L _1/2 sparse modeling can yield solutions more sparse than those of 1 norm and 2 norm, and, furthermore, the model can be efficiently solved by a simple iterative thresholding procedure. The objective function dealt with in L _1/2 regularization theory is, however, of square form, the gradient of which is linear in its variables (such an objective function is the so-called linear gradient function). In this paper, through extending the linear gradient function of L _1/2 regularization framework to the logistic function, we propose a novel sparse version of KLR, the 1/2 quasi-norm kernel logistic regression (1/2-KLR). The version integrates advantages of KLR and L _1/2 regularization, and defines an efficient implementation scheme of sparse KLR. We suggest a fast iterative thresholding algorithm for 1/2-KLR and prove its convergence. We provide a series of simulations to demonstrate that 1/2-KLR can often obtain more sparse solutions than the existing sparsity driven versions of KLR, at the same or better accuracy level. The conclusion is also true even in comparison with sparse SVMs (1-SVM and 2-SVM). We show an exclusive advantage of 1/2-KLR that the regularization parameter in the algorithm can be adaptively set whenever the sparsity (correspondingly, the number of support vectors) is given, which suggests a methodology of comparing sparsity promotion capability of different sparsity driven classifiers. As an illustration of benefits of 1/2-KLR, we give two applications of 1/2-KLR in semi-supervised learning, showing that 1/2-KLR can be successfully applied to the classification tasks in which only a few data are labeled.     

A unified level set framework utilizing parameter priors for medical image segmentation

Image segmentation plays an important role in many medical imaging systems, yet in complex circumstances it remains an open problem. One of the main difficulties is the intensity inhomogeneity in an image. In order to tackle this problem, we first introduce a region-based level set segmentation framework to unify the traditional global and local methods. We then propose two novel parameter priors, i.e., the local order regularization and interactive regularization, and then utilize them as the constraints of the objective energy function. The objective energy function is finally minimized via a level set evolution process to achieve image segmentation. Extensive experiments show that the proposed approach has gained significant improvements in both accuracy and efficiency over the state-of-the-art methods.     

长江中下游护岸生态修复现状与探讨

河道岸坡是水生生态系统与陆地生态系统之间的过渡带,是一个动态的生态系统,具有相互补给水陆之间的水分和能量交换、净化水体、为动植物提供栖息地等极其重要的功能和价值。分析了长江中下游河岸带资源的减少原因和现状各种护岸工程的不足,结合长江中下游实际情况,阐述了河岸带的功能,修复岸坡生态系统的意义,并对长江中下游岸坡修复技术做了系统探讨。     

流域水资源量保护与恢复若干问题的探讨

人口的大量增加,城市的迅速扩张,需水量增加,水资源过度地被开发、水环境恶化,导致人类可以利用的水资源越来越少,水资源保护实为重中之重。流域水资源的保护一度主要侧重于水质的保护,水量保护与恢复没有得到应有的关注。根据水资源量的特征和保护的意义,分析水资源量保护和恢复影响因素,探讨水资源保护的相关问题。     

Development of a new indicator of pollutant loads and its application to the Chesapeake Bay watershed

Pollutant load reductions are often required to restore aquatic ecosystems experiencing eutrophication. Loads can be estimated using watershed models or data from monitoring stations, however data availability can limit the timeliness or comprehensiveness of the load estimates. We developed an approach to address this challenge that used watershed model results to estimate the proportion of annual nonpoint source nitrogen (N), phosphorus (P) and sediment (Sed) loads derived from unmonitored catchments. This proportion was multiplied by the nonpoint portion of United States Geological Survey (USGS) estimated annual river loads to account for annual variation in hydrologic conditions. Total loads were calculated as the sum of measured river loads, reported point sources from unmonitored areas and the estimated nonpoint source loads from unmonitored catchments. We applied this approach to the Chesapeake Bay because of its socio‐economic and ecological importance. Median watershed loads for N, P and Sed were 140, 6.4 and 3030 Mg year^?1, respectively (1990–2004). Nonpoint source loads from the monitored areas constituted the greatest source of N, P and Sed (55, 47 and 74% respectively) to the Bay. The high N, P and Sed yield rates (7.3, 0.38 and 99 kg ha^?1 year^?1, respectively) from nonpoint loads originating from unmonitored areas near the Bay resulted in 25, 32 and 26% (N, P and Sed, respectively) of the Bay's total loads (excluding direct atmospheric deposition, shoreline erosion and oceanic inputs). Disproportionately high loads of P and Seds were associated with years that experienced elevated discharge whereas N loads were directly related to discharge. Error estimates indicated that our methods were most reliable for N (±6%) but reasonable for P (±22%) and provide an effective technique for the timely estimation of pollutant loads from watersheds with unmonitored catchments. Management strategies that decrease N deposition and reduce runoff to control P and Sed transport will effectively reduce pollutant loads. Published in 2010 by John Wiley & Sons, Ltd.     

Geomorphic comparison of two Atlantic coastal rivers: Toward an understanding of physical controls on Atlantic salmon habitat

River channel substrate size and mobility are important to Atlantic salmon spawning and rearing success. We compare morphology and bed sediment between two North American Atlantic coastal streams (Narraguagus River, Maine, USA and Jacquet River, New Brunswick, Canada). The watersheds have similar drainage areas and mean annual precipitation, but differing relief structure, channel longitudinal profiles and numbers of returning salmon. The lower‐relief Narraguagus River is segmented into steeper (gradient >0.002) and flatter reaches (gradient <0.0005). Flat reaches, including mainstem lakes, act as sediment sinks, preventing the continuity of downstream sediment transport. Based on field measurements, the Narraguagus River has a larger high‐flow width to depth ratio than the Jacquet River, but this difference is principally the result of outliers from low‐gradient channel reaches. Measurements of substrate grain size reveal finer river‐bed sediments on the Narraguagus River, however, Shields parameter calculations indicate that bed sediment should be mobile during high flows in both streams. We use the Shields equation to predict grain size based on channel slope, width and drainage area measured from digital elevation models (DEM) and aerial photographs. Predictions of median grain size agree with field measurements within a factor of 2 for 62–70% of the survey stations. We suggest ways that model misfits may provide opportunities to prioritize reach‐based restoration efforts. Based on expected grain size, we estimate 62% spawning and 68% rearing habitat along the length of the Narraguagus River, and 28 and 95% respectively on the Jacquet. Overall, glacial history and relief structure appear to be the first‐order controls on substrate grain size and habitat quality in these two rivers. Copyright © 2010 John Wiley & Sons, Ltd.     

甲骨拓片字形图像复原方法

提出了一种基于自适应阈值和分形几何的甲骨拓片字形图像复原方法。文章分析了甲骨拓片噪声的特点以及字形图像边缘的分形特征,通过计算自适应阈值对噪声区域进行填充。采用统计的方法计算甲骨拓片字形图像边缘的分形维数特征,对字形图像边缘进行压缩变换,进而对甲骨拓片字形图像边缘进行平滑。实验结果显示,这一方法的图像复原效果是比较明显的。     

基于遗传算法和LM优化的BP神经网络的图像复原算法

提出了一种针对图像复原的GA-BP算法.在对退化图像进行复原的过程中,针对BP算法易陷入局部极小、收敛速度慢的缺点,根据遗传算法具有全局寻优的特点,充分利用遗传算法的全局搜索性能进行"粗"搜索,当搜索到全局最优点的附近时,再采用BP算法进行局部搜索,将两者结合起来形成GA-BP算法.通过对图像复原后的效果进行比较,实验表明该算法可以有效的应用于图像复原.     

部分模糊核已知的混合模糊图像复原算法

真实环境中得到的图像常会受到多种模糊降质过程的影响,导致成像质量下降,为此提出一种图像退化模型及相应的复原算法.与传统的级联方式不同,该算法假定模糊核函数是散焦与运动模糊的加权和形式;在散焦模糊给定的情况下,使用广义拉普拉斯分布作为运动模糊的统计模型,并在期望最大化(EM)算法框架下估计模糊核函数;最后利用估计的模糊核函数进行图像复原.实验结果表明,采用文中的复原算法能够较为准确地辨识出模糊核,提升了图像复原的效果.     

改进的两步迭代阈值的遥感图像恢复算法

结合小波变换和Contourlet变换的多尺度、多分辨的共性及这两种变换分别适合处理点奇异和线奇异的特点,提出了一种新的联合使用小波正则项和Contourlet正则项的遥感图像恢复算法。算法中逆问题的求解等价于一个无约束凸规划的求解问题,目标函数由观测图像的拟合优度和正则项组成,传统的正则项是一个函数,本文使用两个正则项函数,能更好地利用图像的先验知识。然后根据Besov空间的半范数等价于小波系数的范数这一原理,提出了基于小波变换的联合使用小波正则项和Contourlet正则项的两步迭代阈值算法。对遥感图像的恢复结果表明,该算法在改善的信噪比（ISNR）和相关系数（CORR）等评价指标上都有显著的改善。