一类新型的光滑支持向量分类机

为了解决支持向量机中非光滑问题,基于贝塞尔函数提出一范式的贝塞尔光滑支持向量机模型,并证明了贝塞尔函数的光滑性和收敛性,分析了其对正号函数的逼进性能。根据模型的特点,应用Armijo-Newton方法进行求解,理论分析和数值实验结果都证明贝塞尔光滑支持向量机在分类性能上优于以往提出来的光滑模型。

Web integration based on classification ontology

In order to eliminate semantic heterogeneity and implement semantic combination in web information integration, the classification ontology is introduced into web information integration. It constructs a standard classification ontology based on web-glossary by extracting classified structures of websites and building mappings between them in order to get unified views. Mapping is defined by calculating concept subordinate matching degrees, concept associate matching degrees and concept dominate matching degrees. A web information integration system is realized, which can effectively solve the problem of classification semantic heterogeneity and implement the integration of web information source and the personal configuration of users.

DIMENSIONALITY REDUCTION BASED ON SVM AND LDA,AND ITS APPLICATION TO CLASSIFICATION TECHNIQUE

Some dimensionality reduction （DR） approaches based on support vector machine （SVM） are proposed. But the acquirement of the projection matrix in these approaches only considers the between-class margin based on SVM while ignoring the within-class information in data. This paper presents a new DR approach, call- ed the dimensionality reduction based on SVM and LDA （DRSL）. DRSL considers the between-class margins from SVM and LDA, and the within-class compactness from LDA to obtain the projection matrix. As a result, DRSL can realize the combination of the between-class and within-class information and fit the between-class and within-class structures in data. Hence, the obtained projection matrix increases the generalization ability of subsequent classification techniques. Experiments applied to classification techniques show the effectiveness of the proposed method.

DISCRIMINATIVE REGULARIZATION:A NEW CLASSIFIER LEARNING METHOD

A novel regularization method -- discriminative regularization （DR）is presented. The method provides a general way to incorporate the prior knowledge for the classification. By introducing the prior information into the regularization term, DR is used to minimize the empirical loss between the desired and actual outputs, as well as maximize the inter-class separability and minimize the intra-class compactness in the output space simultane- ously. Furthermore, by embedding equality constraints in the formulation, the solution of DR can solve a set of linear equations. Classification experiments show the superiority of the proposed DR.

Application of GIS in Ecological Land Type(ELT)mapping--A case in Changbai Mountain area

This paper depicted the physiographic landscape features and natural vegetation situation of study area (the eastern Jilin Province), and expatiates the definition, basic characters and its development of Ecological Land Classification (ELC). Based on the combination of relief map, satellite photography for study area and vegetation inventory data of 480 sample sites, a 5-class and a 15-class ecological land type map was concluded according to 4 important factors including slope, aspect, vegetation and elevation. Ecological Classification System (ECS) is a method to identify, characterize, and map ecosystems. The Ecological Land Type (ELT) was examined and applied initially in eastern Jilin Province.

Classification Method for Dongting Lake Wetland Based on Geographic Information

[Objective] This study aimed to improve the accuracy of remote sensing classification for Dongting Lake Wetland.[Method] Based on the TM data and ground GIS information of Donting Lake,the decision tree classification method was established through the expert classification knowledge base.The images of Dongting Lake wetland were classified into water area,mudflat,protection forest beach,Carem spp beach,Phragmites beach,Carex beach and other water body according to decision tree layers.[Result] The accuracy of decision tree classification reached 80.29%,which was much higher than the traditional method,and the total Kappa coefficient was 0.883 9,indicating that the data accuracy of this method could fulfill the requirements of actual practice.In addition,the image classification results based on knowledge could solve some classification mistakes.[Conclusion] Compared with the traditional method,the decision tree classification based on rules could classify the images by using various conditions,which reduced the data processing time and improved the classification accuracy.

Preoperatively molecular staging with CM10 ProteinChip and SELDI-TOF-MS for colorectal cancer patients

Objectives： To detect the serum proteomic patterns by using SELDI-TOF-MS （surface enhanced laser desorption/ ionization-time of flight-mass spectrometry） technology and CM10 ProteinChip in colorectal cancer （CRC） patients, and to evaluate the significance of the proteomic patterns in the tumour staging of colorectal cancer. Methods： SELDI-TOF-MS and CM10 ProteinChip were used to detect the serum proteomic patterns of 76 patients with colorectal cancer, among them, 10 Stage Ⅰ, 19 Stage Ⅱ, 16 Stage Ⅲ and 31 Stage Ⅳ samples. Different stage models were developed and validated by support vector machines, disctiminant analysis and time-sequence analysis. Results： The Model Ⅰ formed by 6 protein peaks （m/z： 2759.58, 2964.66, 2048.01, 4795.90, 4139.77 and 37761.60） could be used to distinguish local CRC patients （Stage Ⅰ and Stage Ⅱ） from regional CRC patients （Stage Ⅲ） with an accuracy of 86.67% （39/45）. The Model Ⅱ formed by 3 protein peaks （m/z： 6885.30, 2058.32 and 8567,75） could be used to distinguish locoregional CRC patients （Stage Ⅰ, Stage Ⅱ and Stage Ⅲ） from systematic CRC patients （Stage IV） With an accuracy of 75.00% （57/76）. The Model Ⅲ could distinguish Stage Ⅰ from Stage Ⅱ with an accuracy of 86.21% （25/29）. The Model Ⅳ could distinguish Stage Ⅰ from Stage Ⅲ with accuracy of 84.62% （22/26）. The Model Ⅴ could distinguish Stage Ⅱ from Stage Ⅲ with accuracy of 85.71% （30/35）. The Model Ⅵ could distinguish Stage Ⅱ from Stage Ⅳ with accuracy of 80.00% （40/50）. The Model Ⅶ could distinguish Stage Ⅲ from Stage Ⅳ with accuracy of 78.72% （37/47）. Different stage groups could be distinguished by the two-dimensional scattered spots figure obviously. Conclusion： This method showed great success in preoperatively determining the colorectal cancer stage of patients.

Social Network Information Propagation Model Based on Individual Behavior

In this paper, we discuss building an information dissemination model based on individual behavior. We analyze the individual behavior related to information dissemination and the factors that affect the sharing behavior of individuals, and we define and quantify these factors. We consider these factors as characteristic attributes and use a Bayesian classifier to classify individuals. Considering the forwarding delay characteristics of information dissemination, we present a random time generation method that simulates the delay of information dissemination. Given time and other constraints, a user might not look at all the information that his/her friends published. Therefore, this paper proposes an algorithm to predict information visibility, i.e., it estimates the probability that an individual will see the information. Based on the classification of individual behavior and combined with our random time generation and information visibility prediction method, we propose an information dissemination model based on individual behavior. The model can be used to predict the scale and speed of information propagation. We use data sets from Sina Weibo to validate and analyze the prediction methods of the individual behavior and information dissemination model based on individual behavior. A previously proposedinformation dissemination model provides the foundation for a subsequent study on the evolution of the network and social network analysis. Predicting the scale and speed of information dissemination can also be used for public opinion monitoring.

GIS-Based Red Soil Resources Classification andEvaluation

A small scale red soil resources information system (RSRIS) with applied mathematical models wasdeveloped and applied in red soil resources (RSR) classification and evaluation, taking Zhejiang Province,a typical distribution area of red soil, as the study area. Computer-aided overlay was conducted to classifyRSR types. The evaluation was carried out by using three methods, i.e., index summation, square root ofindex multiplication and fuzzy comprehensive assessment, with almost identical results. The result of indexsummation could represent the basic qualitative condition of RSR, that of square root of index multiplicationreflected the real condition of RSR qualitative rank, while fuzzy comprehensive assessment could satisfactorilyhandle the relationship between the evaluation factors and the qualitative rank of RSR, and therefore it is afeasible method for RSR evaluation.

Quantitative information measurement and application for machine component classification codes

Information embodied in machine component classification codes has internal relation with the probability distribu- tion of the code symbol. This paper presents a model considering codes as information source based on Shannon’s information theory. Using information entropy, it preserves the mathematical form and quantitatively measures the information amount of a symbol and a bit in the machine component classification coding system. It also gets the maximum value of information amount and the corresponding coding scheme when the category of symbols is fixed. Samples are given to show how to evaluate the information amount of component codes and how to optimize a coding system.

Characterizing Internet Backbone Traffic Based on Deep Packets Inpection and Deep Flows Inspection

Based on the massive data collected with a passive network monitoring equipment placed in China＇s backbone, we present a deep insight into the network backbone traffic and evaluate various ways for inproving traffic classifying efficiency in this pa- per. In particular, the study has scrutinized the net- work traffic in terms of protocol types and signatures, flow length, and port distffoution, from which mean- ingful and interesting insights on the current Intemet of China from the perspective of both the packet and flow levels are derived. We show that the classifica- tion efficiency can be greatly irrproved by using the information of preferred ports of the network applica- tions. Quantitatively, we find two traffic duration thresholds, with which 40% of TCP flows and 70% of UDP flows can be excluded from classification pro- cessing while the in^act on classification accuracy is trivial, i.e., the classification accuracy can still reach a high level by saving 85% of the resources.

A SPATIAL CLUSTER METHOD SUPPORTED BY GIS FOR URBAN-SUBURBAN-RURAL CLASSIFICATION

This study was undertaken to construct a preliminary spatial analysis method for building an urban-suburban-rural category in the specific sample area of central California and providing distribution characteristics in each category, based on which, some further studies such as regional manners of residential wood burning emission (PM2.5, the term used for a mixture of solid particles and liquid droplets found in the air, refers to particulate matter that is 2.5 mu m or smaller in size) could be carried out for the project of residential wood combustion. Demographic and infrastructure data with spatial characteristics were processed by integrating both Geographic Information System (GIS) and statistics method (Cluster Analysis), and then output to a category map as the result. It approached the quantitative and multi-variables description on the major characteristics variations among the urban, suburban and rural; and perfected the TIGER's urban-rural classification scheme by adding suburban category. Based on the free public GIS data, the spatial analysis method provides an easy and ideal tool for geographic researchers, environmental planners, urban/regional planners and administrators to delineate different categories of regional function on the specific locations and dig out spatial distribution information they wanted. Furthermore, it allows for future adjustment on some parameters as the spatial analysis method is implemented in the different regions or various eco-social models.

A Method of Soil Salinization Information Extraction with SVM Classification Based on ICA and Texture Features

Salt-affected soils classification using remotely sensed images is one of the most common applications in remote sensing,and many algorithms have been developed and applied for this purpose in the literature.This study takes the Delta Oasis of Weigan and Kuqa Rivers as a study area and discusses the prediction of soil salinization from ETM +Landsat data.It reports the Support Vector Machine(SVM) classification method based on Independent Component Analysis(ICA) and Texture features.Meanwhile,the letter introduces the fundamental theory of SVM algorithm and ICA,and then incorporates ICA and texture features.The classification result is compared with ICA-SVM classification,single data source SVM classification,maximum likelihood classification(MLC) and neural network classification qualitatively and quantitatively.The result shows that this method can effectively solve the problem of low accuracy and fracture classification result in single data source classification.It has high spread ability toward higher array input.The overall accuracy is 98.64%,which increases by10.2% compared with maximum likelihood classification,even increases by 12.94% compared with neural net classification,and thus acquires good effectiveness.Therefore,the classification method based on SVM and incorporating the ICA and texture features can be adapted to RS image classification and monitoring of soil salinization.

Earnings disaggregation and analysts＇ forecasts

Accounting concepts dictate that separately disclosed components should contain separate useful information. This paper examines the relations between income statement components and analysts＇ earnings forecasts and forecast errors. Regressions explaining earnings forecasts using earnings components provide a better fit than regression using just aggregate income to explain forecasts. We interpret this as consistent with the hypothesis that analysts use incremental information in components not available in aggregate income. However, additional tests based on predictability of forecast errors indicate that analysts do not incorporate all information available in components into earnings forecasts. In addition, this inefficiency appears to increase at longer forecast horizons.

Semantic categorization of indoor places using CNN for mobile robot exploration

The ability of achieving a semantic understanding of workspaces is an important capability for mobile robot. A method is proposed to categorize different places in a typical indoor environment by using a Kinect sensors for mobile robot exploration. At first, the invariant feature based images stitching approach is adopted to form a panoramic image according to Kinect visual information, and the translation between Kinect depth information and obstacle distance information is performed to obtain virtual LIDAR data. Then, the semantic classifier is designed by using convolutional neural networks （CNN） for indoor place eategorization based on Kinect visual observations with panoramic view. At last, a frontier-based exploration method is applied to carry out indoor autonomous exploration of mo- bile robots, which integrates the CNN-based categorization approach. The proposed method has been implemented and tested on a real robot, and experiment results demonstrate the approach effective- ness on solving the semantic categorization problem for mobile robot exploration.

Quantum Information Splitting of an Arbitrary Three-Qubit State by Using Cluster States and Bell States

A new application of cluster states is investigated for quantum information splitting （QIS） of an arbitrary three-qubit state. In our scheme, a four-qubit cluster state and a Bell state are shared by a sender （Alice）, a controller （Charlie）, and a receiver （Bob）. Both the sender and controller only need to perform Bell-state measurements （BSMs）, the receiver can reconstruct the arbitrary three-qubit state by performing some appropriately unitary transformations on his qubits after he knows the measured results of both the sender and the controller. This QIS scheme is deterministic.

A Geographic Information Systems approach for classifying and mapping forest management category in Baihe Forestry Bureau, Northeast China

This paper demonstrates a Geographic Information Systems （GIS） procedure of classifying and mapping forest management category in Baihe Forestry Burea, Jilin Province, China. Within the study area, Baihe Forestry Bureau land was classified into a two-hierarchy system. The top-level class included the non-forest and forest. Over 96% of land area is forest in the study area, which was further divided into key ecological service forest （KES）, general ecological service forest （GES）, and commodity forest （COM）. COM covered 45.0% of the total land area and was the major forest management type in Baihe Forest Bureau. KES and GES accounted for 21.2% and 29.9% of the total land area, respectively. The forest management zones designed with GIS in this study were then compared with the forest management zones established using the hand draw by the local agency. There were obvious differences between the two products. It suggested that the differences had some to do with the data sources, basic unit and mapping procedures. It also suggested that the GIS method was a useful tool in integrating forest inventory data and other data for classifying and mapping forest zones to meet the needs of the classified forest management system.

Landslide hazards mapping using uncertain Na?ve Bayesian classification method

Landslide hazard mapping is a fundamental tool for disaster management activities in Loess terrains. Aiming at major issues with these landslide hazard assessment methods based on Naive Bayesian classification technique, which is difficult in quantifying those uncertain triggering factors, the main purpose of this work is to evaluate the predictive power of landslide spatial models based on uncertain Naive Bayesian classification method in Baota district of Yan＇an city in Shaanxi province, China. Firstly, thematic maps representing various factors that are related to landslide activity were generated. Secondly, by using field data and GIS techniques, a landslide hazard map was performed. To improve the accuracy of the resulting landslide hazard map, the strategies were designed, which quantified the uncertain triggering factor to design landslide spatial models based on uncertain Naive Bayesian classification method named NBU algorithm. The accuracies of the area under relative operating characteristics curves（AUC） in NBU and Naive Bayesian algorithm are 87.29% and 82.47% respectively. Thus, NBU algorithm can be used efficiently for landslide hazard analysis and might be widely used for the prediction of various spatial events based on uncertain classification technique.

Design and implementation of a large-scale multi-class text classifier

Although, researchers in the ATC field have done a wide range of work based on SVM, almost all existing approaches utilize an empirical model of selection algorithms. Their attempts to model automatic selection in practical, large-scale, text classification systems have been limited. In this paper, we propose a new model selection algorithm that utilizes the DDAG learning architecture. This architecture derives a new large-scale text classifier with very good performance. Experimental results show that the proposed algorithm has good efficiency and the necessary generalization capability while handling large-scale multi-class text classification tasks.

Steganalysis of LSB Matching Using Characteristic Function Moment of Pixel Differences

Nowadays,many steganographic tools have been developed,and secret messages can be imperceptibly transmitted through public networks.This paper concentrates on steganalysis against spatial least significant bit(LSB) matching,which is the prototype of many advanced information hiding methods.Many existing algorithms deal with steganalysis problems by using the dependencies between adjacent pixels.From another aspect,this paper calculates the differences among pixel pairs and proves that the histogram of difference values will be smoothed by stego noises.We calculate the difference histogram characteristic function(DHCF) and deduce that the moment of DHCFs(DHCFM) will be diminished after stego bits are hidden in the image.Accordingly,we compute the DHCFMs as the discriminative features.We calibrate the features by decreasing the influence of image content on them and train support vector machine classifiers based on the calibrated features.Experimental results demonstrate that the DHCFMs calculated with nonadjacent pixels are helpful to detect stego messages hidden by LSB matching.