Semisupervised Dimensionality Reduction and Classification Through Virtual Label Regression

Semisupervised dimensionality reduction has been attracting much attention as it not only utilizes both labeled and unlabeled data simultaneously, but also works well in the situation of out-of-sample. This paper proposes an effective approach of semisupervised dimensionality reduction through label propagation and label regression. Different from previous efforts, the new approach propagates the label information from labeled to unlabeled data with a well-designed mechanism of random walks, in which outliers are effectively detected and the obtained virtual labels of unlabeled data can be well encoded in a weighted regression model. These virtual labels are thereafter regressed with a linear model to calculate the projection matrix for dimensionality reduction. By this means, when the manifold or the clustering assumption of data is satisfied, the labels of labeled data can be correctly propagated to the unlabeled data; and thus, the proposed approach utilizes the labeled and the unlabeled data more effectively than previous work. Experimental results are carried out upon several databases, and the advantage of the new approach is well demonstrated.

中国水蚀土壤抗剪强度研究

利用野外测试数据、土壤样品的室内理化分析数据和参考文献数据 ,对水蚀区范围内水土流失过程中的土壤抗剪强度进行了初步研究 ,建立了中国水土流失土壤抗剪强度的回归模型 ,总结出水蚀区范围内水蚀过程中有关土壤抗剪强度的 3条结论 :影响水土流失过程中土壤抗剪强度的主导因素是容重、粉 /黏、土壤含水量、土壤有机质含量 ;抗剪强度随土壤类型发生有规律的变化 ;抗剪强度在中国水蚀区范围内有较明显的空间分异规律 (包括水平分异规律和垂直剖面构型规律 )

A field study on the conversion ratio of phytoplankton biomass carbon to chlorophyll-a in Jiaozhou Bay, China

A one-year field study was conducted to determine the conversion ratio of phytoplankton biomass carbon (Phyto-C) to chlorophyll-a (Chl-a) in Jiaozhou Bay, China. We measured suspended particulate organic carbon (POC) and phytoplankton Chl-a samples collected in surface water monthly from March 2005 to February 2006. The temporal and spatial variations of Chl-a and POC concentrations were observed in the bay. Based on the field measurements, a linear regression model II was used to generate the conversion ratio of Phyto-C to Chl-a. In most cases, a good linear correlation was found between the observed POC and Chl-a concentrations, and the calculated conversion ratios ranged from 26 to 250 with a mean value of 56 A mu g A mu g(-1). The conversion ratio in the fall was higher than that in the winter and spring months, and had the lowest values in the summer. The ratios also exhibited spatial variations, generally with low values in the near shore regions and relatively high values in offshore waters. Our study suggests that temperature was likely to be the main factor influencing the observed seasonal variations of conversion ratios while nutrient supply and light penetration played important roles in controlling the spatial variations.

装配系统中缓冲区容量的研究

为了确定装配系统中的缓冲区容量,在建立缓冲区状态数学模型的基础上,根据随机过程的原理,提出了缓冲区被充满概率和缓冲区容量之间的函数关系。以缓冲区被充满概率最小化为目标,确定合理的缓冲区容量。最后给出一种递进算法,通过回归方程计算缓冲区对装配工位生产率的影响,逐步求出由多个工位组成的整个装配系统各个工位之间的缓冲区容量。

地理信息系统支持下的滑坡灾害空间分析预测-以云南小江流域为例

As a typical geological and environmental hazard, landslide has been causing more and more property and life losses. However, to predict its accurate occurring time is very difficult or even impossible due to landslide's complex nature. It has been realized that it is not a good solution to spend a lot of money to treat with and prevent landslide. The research trend is to study landslide's spatial distribution and predict its potential hazard zone under certain region and certain conditions. GIS(Geographical Information System) is a power tools for data management, spatial analysis based on reasonable spatial models and visualization. It is new and potential study field to do landslide hazard analysis and prediction based on GIS. This paper systematically studies the theory and methods for GIS based landslide hazard analysis. On the basis of project "Mountainous hazard study-landslide and debris flows" supported by Chinese Academy of Sciences and the former study foundation, this paper carries out model research, application, verification and model result analysis. The occurrence of landslide has its triggering factors. Landslide has its special landform and topographical feature which can be identify from field work and remote sensing image (aerial photo). Historical record of landslide is the key to predict the future behaviors of landslide. These are bases for landslide spatial data base construction. Based on the plenty of literatures reviews, the concept framework of model integration and unit combinations is formed. Two types of model, CF multiple regression model and landslide stability and hydrological distribution coupled model are bought forward. CF multiple regression model comes form statistics and possibility theory based on data. Data itself contains the uncertainty and random nature of landslide hazard, so it can be seen as a good method to study and understand landslide's complex feature and mechanics. CF multiple regression model integrates CF (landslide Certainty Factor) and multiple regression prediction model. CF can easily treat with the problems of data quantifying and combination of heteroecious data types. The combination of CF can assist to determine key landslide triggering factors which are then inputted into multiple regression model. CF regression model can provide better prediction results than traditional model. The process of landslide can be described and modeled by suitable physical and mechanical model. Landslide stability and hydrological distribution coupled model is such a physical deterministic model that can be easily used for landslide hazard analysis and prediction. It couples the general limit equilibrium method and hydrological distribution model based on DEM, and can be used as a effective approach to predict the occurrence of landslide under different precipitation conditions as well as landslide mechanics research. It can not only explain pre-existed landslides, but also predict the potential hazard region with environmental conditions changes. Finally, this paper carries out landslide hazard analysis and prediction in Yunnan Xiaojiang watershed, including landslide hazard sensitivity analysis and regression prediction model based on selected key factors, determining the relationship between landslide occurrence possibility and triggering factors. The result of landslide hazard analysis and prediction by coupled model is discussed in details. On the basis of model verification and validation, the modeling results are showing high accuracy and good applying potential in landslide research.

Prediction of soil adsorption coefficients from retention parameters on three reversed-phase liquid chromatographic columns

Reversed-phase high-performance liquid chromatographic (RP-HPLC) retention parameters, which are determined by the intermolecular interactions in retention process, can be considered as the chemical molecular descriptors in linear free energy relationships (LFERs). On the basis of the characterization and comparison of octadecyl-bonded silica gel (ODS), cyano-bonded silica gel (CN), and phenyl-bonded silica gel (Ph) columns with linear solvation energy relationships (LSERs), a new multiple linear regression model using RP-HPLC retention parameters on ODS and CN columns as variables for estimation of soil adsorption coefficients was developed. It was tested on a set of reference substances from various chemical classes. The results showed that the multicolumn method was more promising than a single-column method was for the estimation of soil adsorption coefficients. The accuracy of the suggested model is identical with that of LSERs.

对癌症生存者应对资源的研究

A survey study of cancer survivors was conducted to explore the coping resources, which buffers the life of cancer survivors against stressful situation. Participants reported coping strategies, positive affect and negative affect, personality, perceived social support, fighting spirit and helpless/hopeless as well as quality of life through a set of self-assessment questionnaire. The results indicated that the frequency of coping strategies used by cancer survivors from high to low were: growing, problem solving, seeking support,self-controlling, wishful thinking, and distancing. The correlational analysis indicated that among the six sets of coping strategies, growing was positively correlated most strongly with most of the dimensions in quality of life as well as positive affect. Among the five personality, Neuroticism was positively correlated most strongly with helpless/hopeless and negative affect; and was negatively correlated most strongly with fighting spirit and positive affect. Extraversion was positively correlated most strongly with positive affect and negatively correlated most strongly with helpless/hopeless; Agreeableness was negatively correlated most strongly with negative affect; Conscientiousness was positively correlated most strongly with fighting spirit. Subjects with higher score in quality of life reported higher frequency of coping strategies in growing and problem solving and less in wishful thinking. They also reported higher scores in Extraversion, Agreeableness, Conscientiousness as well as lower scores in Neuroticism. The regression analysis displayed that not negative affect but positive affect entered the regression model when all the psychological and social variables in the study were accounted for. Taken together, these data suggested that, growing was the most effective coping strategy among the six sets of strategies for cancer survivors to improve quality of life, to maintain positive affect and to enhance fighting spirit. Neuroticism was vulnerable to resist stressors; Extraversion, Agreeableness, and Conscientiousness were stress-resisted factors. Positive affect may has more adaptational significance than negative affect during chronic stress. These data also implicated that positive affect should be paid more attention to in coping research.