Search

Network embedding aims to learn a low-dimensional vector for each node in networks, which is effective in a variety of applications such as network reconstruction and community detection. However, the majority of the existing network embedding methods merely exploit the network structure and ignore the rich node attributes, which tend to generate sub-optimal network representation. To learn more desired network representation, diverse information of networks should be exploited. In this paper, we develop a novel deep autoencoder framework to fuse topological structure and node attributes named FSADA. We firstly design a multi-layer autoencoder which consists of multiple non-linear functions to capture and preserve the highly non-linear network structure and node attribute information. Particularly, we adopt a pre-processing procedure to pre-process the original information, which can better facilitate to extract the intrinsic correlations between topological structure and node attributes. In addition, we design an enhancement module that combines topology and node attribute similarity to construct pairwise constraints on nodes, and then a graph regularization is introduced into the framework to enhance the representation in the latent space. Our extensive experimental evaluations demonstrate the superior performance of the proposed method.

This paper proposes a network anomaly detection model of direct batch growing hierarchical self-organizing mapping based on entropy, which facilitates clear topology representation for the asymmetrically-distributed data. Since the entropy-defined parameters dynamically vary with the incident dataset, that is, follow a data-adaptive manner, the proposed model is naturally valid in all cases with various data types. For fine-grained data distinguishing, a resemble entropy parameter is proposed for the first time to our best knowledge. The experimental results validate that the proposed model achieves a more efficient network anomaly detection than the conventional models, especially for real-world applications with unexpected anomaly data updating.

Network embedding is an effective method aiming to learn the low-dimensional vector representation of nodes in networks, which has been widely used in various network analytic tasks such as node classification, node clustering, and link prediction. The objective of network embedding is to capture the structural information and inherent characteristics of the network as much as possible in the low-dimensional vector representation. However, the majority of the existing network embedding methods merely exploited the microscopic proximity of the network structure to learn the node representation, which tend to generate sub-optimal network representation. In this paper, we propose a novel nonnegative matrix factorization (NMF) based network representation learning framework called FLGAI, which jointly integrates the local network structure, global network structure, and attribute information to learn the network representation. First, we employ the first-order proximity and second-order proximity jointly to preserve the local network structure. Then, the community structure is introduced to preserve the global network structure. Third, we exploit the node attribute information to capture the node characteristics. To preserve the structural information and the network node attributes simultaneously, we formulate their consensus relationships and optimize them jointly in a unified NMF framework to derive the final network representation. To evaluate the effectiveness of our model, we conduct extensive experiments on six real-world datasets and the empirical results demonstrate the superior performance of the proposed method over the state-of-the-art approaches in both node classification and node clustering tasks.

Applying an accurate parametric prediction model to identify abnormal or false pressurizer water levels (PWLs) is critical to the safe operation of marine pressurized water reactors (PWRs). Recently, deep-learning-based models have proved to be a powerful feature extractor to perform high-accuracy prediction. However, the effectiveness of models still suffers from two issues in PWL prediction: the correlations shifting over time between PWL and other feature parameters, and the example imbalance between fluctuation examples (minority) and stable examples (majority). To address these problems, we propose a cost-sensitive mechanism to facilitate the model to learn the feature representation of later examples and fluctuation examples. By weighting the standard mean square error loss with a cost-sensitive factor, we develop a Cost-Sensitive Long Short-Term Memory (CSLSTM) model to predict the PWL of PWRs. The overall performance of the CSLSTM is assessed by a variety of evaluation metrics with the experimental data collected from a marine PWR simulator. The comparisons with the Long Short-Term Memory (LSTM) model and the Support Vector Regression (SVR) model demonstrate the effectiveness of the CSLSTM.

Background subtraction is a classic video processing task pervading in numerous visual applications such as video surveillance and traffic monitoring. Given the diversity and variability of real application scenes, an ideal background subtraction model should be robust to various scenarios. Even though deep-learning approaches have demonstrated unprecedented improvements, they often fail to generalize to unseen scenarios, thereby less suitable for extensive deployment. In this work, we propose to tackle cross-scene background subtraction via a two-phase framework that includes meta-knowledge learning and domain adaptation. Specifically, as we observe that meta-knowledge (i.e., scene-independent common knowledge) is the cornerstone for generalizing to unseen scenes, we draw on traditional frame differencing algorithms and design a deep difference network (DDN) to encode meta-knowledge especially temporal change knowledge from various cross-scene data (source domain) without intermittent foreground motion pattern. In addition, we explore a self-training domain adaptation strategy based on iterative evolution. With iteratively updated pseudo-labels, the DDN is continuously fine-tuned and evolves progressively toward unseen scenes (target domain) in an unsupervised fashion. Our framework could be easily deployed on unseen scenes without relying on their annotations. As evidenced by our experiments on the CDnet2014 dataset, it brings a significant improvement to background subtraction. Our method has a favorable processing speed (70 fps) and outperforms the best unsupervised algorithm and top supervised algorithm designed for unseen scenes by 9% and 3%, respectively.

Although person re-identification (ReID) has drawn increasing research attention due to its potential to address the problem of analysis and processing of massive monitoring data, it is very challenging to learn discriminative information when the people in the images are occluded, in large pose variations or from different perspectives. To address this problem, we propose a novel joint attention person ReID (JA-ReID) architecture. The idea is to learn two complementary feature representations by combining a soft pixel-level attention mechanism and a hard region-level attention mechanism. The soft pixel-level attention mechanism learns a discriminative embedding for the fine-grained information by exploring the salient parts in the feature maps. The hard region-level attention mechanism conducts uniform partitions on the convolutional feature maps for learning local features. We have achieved competitive results in three popular benchmarks, including Market1501, DukeMTMC-reID, and CUHK03. The experimental results verify the adaptability of the joint attention mechanism to non-rigid deformation of the human body, which can effectively improve the accuracy of ReID.

With the rapid progress of urbanization, predicting citywide crowd flows has become increasingly significant in many fields, such as traffic management and public security. However, influenced by the complex spatiotemporal relations in raw data and other factors, such as events and weather, obtaining a precise prediction is challenging. Some previous works attempted to address this problem using various ways, such as autoregressive integrated moving average, vector auto-regression and some deep learning models. However, seldom can these methods comprehensively capture the spatiotemporal correlations. In this paper, we propose a novel spatio-temporal prediction model that is based on densely connected convolutional networks and attention long short-term memory (ST-DCCNAL), to simultaneously predict the inflow and outflow of the crowds in regions divided within a specific city. The ST-DCCNAL model consists of three parts: spatial part, external factors part and temporal part. In the spatial part, we employ densely connected convolutional networks to extract spatial characteristics at different levels. The external factors part utilizes a fully connected network to extract features from auxiliary information. In the last part, an attention-based long short-term memory module is leveraged to capture the temporal pattern. To demonstrate the practicality and effectiveness of the proposed model, we evaluate it using two separate real-world datasets of taxis in Beijing and bikes in New York. The experimental results confirm that the performance of our model is better than that of other baseline methods.

Malware and its variants continue to pose a threat to network security. Machine learning has been widely used in the field of malware classification, but some emerging studies, such as attention mechanisms, are rarely applied in this field. In this paper, we analyze the correspondence between bytecode and disassembly of malware, and propose a new feature extraction method based on multi-dimensional sequence. Also, we construct a new classification framework based on attention mechanism and Convolutional Neural Networks mechanism. Furthermore, we also compare the different architectures based on the attention mechanisms. Experiments on open datasets show that our feature extraction method and our framework have a good classification effect, and the accuracy rate is 0.9609.

As the network goes deep into all aspects of people''s lives, the number and the complexity of network traffic is increasing, and traffic classification becomes more and more important. How to classify them effectively is an important prerequisite for network management and planning, and ensuring network security. With the continuous development of deep learning, more and more traffic classification begins to use it as the main method, which achieves better results than traditional classification methods. In this paper, we provide a comprehensive review of network traffic classification based on deep learning. Firstly, we introduce the research background and progress of network traffic classification. Then, we summarize and compare traffic classification based on deep learning such as stack autoencoder, one-dimensional convolution neural network, two-dimensional convolution neural network, three-dimensional convolution neural network, long short-term memory network and Deep Belief Networks. In addition, we compare traffic classification based on deep learning with other methods such as based on port number, deep packets detection and machine learning. Finally, the future research directions of network traffic classification based on deep learning are prospected.

Community detection is a classic and challenging network analysis task. Inspired by the similarity between autoencoder and modularity maximization model in terms of a low-dimensional approximation of the modularity matrix. In this paper, a novel community detection framework via deep autoencoder was proposed. Firstly, the modularity is fed into the autoencoder, which can obtain a non-linear deep representation for the network. In addition, node attribute similarity was utilized to construct pairwise constraints on nodes, and then a graph regularization is introduced into the framework. The extensive experimental evaluations on real-world networks and synthetic networks demonstrate that the proposed framework achieves superior performance compared with the state-of-the-art community detection algorithms.

We introduce a three stage pipeline: resized-diverse-inputs (RDIM), diversity-ensemble (DEM) and region fitting, that work together to generate transferable adversarial examples. We first explore the internal relationship between existing attacks, and propose RDIM that is capable of exploiting this relationship. Then we propose DEM, the multi-scale version of RDIM, to generate multi-scale gradients. After the first two steps we transform value fitting into region fitting across iterations. RDIM and region fitting do not require extra running time and these three steps can be well integrated into other attacks. Our best attack fools six black-box defenses with a 93% success rate on average, which is higher than the state-of-the-art gradient-based attacks. Besides, we rethink existing attacks rather than simply stacking new methods on the old ones to get better performance. It is expected that our findings will serve as the beginning of exploring the internal relationship between attack methods.

Visible-infrared person re-identification (VI-ReID) is a significant technology in night-time surveillance applications. Compared to traditional person re-identification that focuses on only visible imaging system, the modality discrepancy between infrared and visible images brings an additional challenge to VI-ReID. Taking the shortcomings of existing VI-ReID works in terms of feature representation into account, in this paper, we propose an end-to-end network named Joint Feature Learning Network (JFLN) to jointly utilize global and local features, high-level and middle-level features, as well as shared and specific features. In addition, based on these extracted features, we introduce a multi-loss supervision strategy in the training process. Extensive experiments conducted on public SYSU-MM01 dataset demonstrate that the proposed method outperforms the state-of-the-arts, with relatively less computational cost.

Canonical correlation analysis (CCA) is one of the most well-known methods to extract features from multi-view data and has attracted much attention in recent years. However, classical CCA is unsupervised and does not take discriminant information into account. In this paper, we add discriminant information into CCA by using random cross-view correlations between within-class samples and propose a new method for multi-view dimensionality reduction called canonical random correlation analysis (RCA). In RCA, two approaches for randomly generating cross-view correlation samples are developed on the basis of bootstrap technique. Furthermore, kernel RCA (KRCA) is proposed to extract nonlinear correlations between different views. Experiments on several multi-view data sets show the effectiveness of the proposed methods.

Malicious software (Malware) has been growing exponentially recently. And it is urgently need to identify Malware from normal software. As Malware samples can be represented as byteplot grayscale images. Thus, deep transfer learning is a proper method for Malware images classification. However, the Malware samples do not possess the vertical direction structure as image. It is more like one dimensional signal and we can convert Malware to a Malware signal. Thus, we try to extract some feature vector from Malware signal. Then we can use SVM to train a model to detect Malware file. Recently, Markov Transition Field (MTF) can encode time series as images for visual inspection and classification. We can try to adapt MTF to encode the Malware signal to MTF image. Then we reshape MTF as a one-dimensional vector for SVM to train a model. Our proposed method is better than the byteplot grayscale image based method in our dataset. And our proposed MTF based two classes Malware detection method can be used for Malware software detection in reality.

Counting the number of person has received much attention in recent years. Most of the existing crowd counting methods adopted density map regression pipeline, which formulates the crowd counting problem to two fragmented part: density map regression and integration of the overall counting. To solve this problem, this paper presents a multi-task deep learning scheme to enhance the counting performance. More specifically, we firstly build a multi-scale deep convolutional neural network, based on combining the feature maps of conv layers with different filters, to solve the multi-scale problem in crowd counting. Secondly, we develop the multi-task structure that can simultaneously learn the density map and the global counting. Experiments on large scale crowd counting datasets, Shanghaitech and WorldExpo10, demonstrate that the proposed method achieves much reduction in counting error respectively

Stock price prediction is an important issue in the financial world, as it contributes to the development of effective strategies for stock exchange transactions. In this paper, we propose a generic framework employing Long Short-Term Memory (LSTM) and convolutional neural network (CNN) for adversarial training to forecast high-frequency stock market. This model takes the publicly available index provided by trading software as input to avoid complex financial theory research and difficult technical analysis, which provides the convenience for the ordinary trader of nonfinancial specialty. Our study simulates the trading mode of the actual trader and uses the method of rolling partition training set and testing set to analyze the effect of the model update cycle on the prediction performance. Extensive experiments show that our proposed approach can effectively improve stock price direction prediction accuracy and reduce forecast error.

Identifying influential nodes is an important topic in many diverse applications, such as accelerating information propagation, controlling rumors and diseases. Many methods have been put forward to identify influential nodes in complex networks, ranging from node centrality to diffusion-based processes. However, most of the previous studies do not take into account overlapping communities in networks. In this paper, we propose an effective method based on network representation learning. The method considers not only the overlapping communities in networks, but also the network structure. Experiments on real-world networks show that the proposed method outperforms many benchmark algorithms and can be used in large-scale networks.

Social network is a hot issue in recent years. In this field, data with multiple views or data from multiple sources are referred to as multi-view data. Due to the uncertain quality of data source, single-view method often leads to unstable performance in community discovery. However, to combine various numbers of views to improve community detection performance is a challenge. In this paper, we propose a method called CoNMTF (Co-regularized Nonnegative Matrix Tri-Factorization). A relaxed pairwise regularization is introduced to integrate multiview adjacency data. Under this framework, we propose an iterative algorithm and prove its correctness and convergence. Experimental results on both synthetic datasets and real-world datasets demonstrate that it outperforms the-state-of-art algorithms in terms of accuracy and NMI.

Machine learning has been used in network traffic classification and statistical features are used to represent flows. However, conventional feature selection may work out in face of dynamic and complex traffic data. Multi-Task Learning has obtained quite wide attention nowadays, and one important form of multi-task learning is to exploit the features shared by tasks by sparse models. We propose a fast multi-task sparse feature learning method, using a non-convex Capped- ? 1 , ? 1 as the regularizer to learn a set of shared features in traffic data. Specifically, the non-convex multi-task feature learning model can learn features belonging to each task as well as the common features shared among tasks. We use the iterative shrinkage and thresholding (IST) algorithm to solve the problem, which has a closed-form solution for one of the crucial steps in the whole iteration. Experiment on real traffic data captured from backbone network as well as synthetic data and other popular real-world data show the effectiveness the method, compared with state-of-the-art methods.

Pairwise constraints, which specifies whether a pair of samples belong to the same class (must-link constraints) or different classes (cannot-link constraints). In this paper, we propose a new feature selection algorithm based on pairwise constraints for linked social media data, which integrate the linking information and pairwise constraints simultaneously to select robust features. As it known to all, the wide use of social media produces massive, high-dimensional and unlabeled social media data. Feature selection has been shown effective in dealing with high-dimensional data. The experiments on Flickr and BlogCatalog datasets evaluate that our method outperforms other conventional feature selection methods.

It is important to identify encrypted data stream on the network. The accurate identification of encrypted data stream not only helps insight into the operation of the entire network, but also accurately controls user behavior for specific requirements. In this paper, we proposed Gaussian mixture model using sparse feature selection of randomness to solve the identification of encrypted data stream. Experimental results show that the average identification rate of encrypted data stream is over 90%.

Background Active modules are connected regions in biological network which show significant changes in expression over particular conditions. The identification of such modules is important since it may reveal the regulatory and signaling mechanisms that associate with a given cellular response. Results In this paper, we propose a novel active module identification algorithm based on a memetic algorithm. We propose a novel encoding/decoding scheme to ensure the connectedness of the identified active modules. Based on the scheme, we also design and incorporate a local search operator into the memetic algorithm to improve its performance. Conclusion The effectiveness of proposed algorithm is validated on both small and large protein interaction networks.

Community detection is an important tasks across a number of research fields including social science, biology, and physics. In the real world, topology information alone is often inadequate to accurately find out community structure due to its sparsity and noise. The potential useful prior information such as pairwise constraints which contain must-link and cannot-link constraints can be obtained from domain knowledge in many applications. Thus, combining network topology with prior information to improve the community detection accuracy is promising. Previous methods mainly utilize the must-link constraints while cannot make full use of cannot-link constraints. In this paper, we propose a semi-supervised community detection framework which can effectively incorporate two types of pairwise constraints into the detection process. Particularly, must-link and cannot-link constraints are represented as positive and negative links, and we encode them by adding different graph regularization terms to penalize closeness of the nodes. Experiments on multiple real-world datasets show that the proposed framework significantly improves the accuracy of community detection.

This paper proposes an accurate crowd counting method based on convolutional neural network and low-rank and sparse structure. To this end, we firstly propose an effective deep-fusion convolutional neural network to promote the density map regression accuracy. Furthermore, we figure out that most of the existing CNN based crowd counting methods obtain overall counting by direct integral of estimated density map, which limits the accuracy of counting. Instead of direct integral, we adopt a regression method based on low-rank and sparse penalty to promote accuracy of the projection from density map to global counting. Experiments demonstrate the importance of such regression process on promoting the crowd counting performance. The proposed low-rank and sparse based deep-fusion convolutional neural network (LFCNN) outperforms existing crowd counting methods and achieves the state-of-the-art performance.

Social networks are widespread and important in our daily life. Finding communities and reveal node characteristics in community are crucial to understand the network structure and function. Many methods based on Nonnegative matrix factorization NMF are proposed to find communities, while these results appear uncertain with the initial condition especially in weighted directed network. In this paper, firstly we improve the nonnegative matrix factorization NMF method with modeling network as the weighted directed graph and using diagonally dominant matrix as constraint condition to obtain the community membership of each node as well as the interaction among communities. Furthermore, we raise methods to evaluate nodes importance and to discuss node characteristics in community to analyze the network structure. Some experiments on the Zachary club datasets and other real-world datasets have been did to demonstrate the superiority of our methods for community discovery over other related matrix factorization methods. The results demonstrate that our methods are useful and applicable both in weighted directed model and undirected model for community discovery, and the results are more reliable. Experiments also illustrate the meaningful results by discussing the node characteristics in community. All those provide a useful way for analyzing social network.

The accurate identification of encrypted data stream helps to regulate illegal data, detect network attacks and protect users' information. In this paper, a novel encrypted data stream identification algorithm is introduced. The proposed method is based on randomness characteristics of encrypted data stream. We use a l1-norm regularized logistic regression to improve sparse representation of randomness features and Fuzzy Gaussian Mixture Model (FGMM) to improve identification accuracy. Experimental results demonstrate that the method can be adopted as an effective technique for encrypted data stream identification.

Multi-label learning is popular in current research of machine learning areas, and there have already been many methods using label relationship to solve multi-label problems. However, the meaning of their relationship is not so obvious that it's hard for us to know the fact among labels. Besides, with the development of multi-label learning, hierarchical multi-label classification is a new research hotspot in the field of machine learning and data mining, that these labels have a hierarchical structure, so it is more and more essential to analyze the structure of them to help us learn the deep meaning of labels. In this paper, we proposed a community discovery approach based on label relationship and NMF algorithm both in common multi-label datasets and hierarchical datasets. With this method, we can find the clear structure in multiple labels, which is helpful for obtaining the label community and catching sight of regulation these tags. After all, it's of great importance for the study of multi-label relationship learning.

Feature extraction is an essential preprocessing step in machine learning and data mining. Generally, supervised feature extraction algorithms with prior knowledge outperform unsupervised ones without prior knowledge. In particular, nearly all existing supervised feature extraction algorithms employ class labels or pairwise constraints as supervised information. In this paper, we propose to employ another form of supervised information, i.e. Universum, which demonstrates a collection of “non-examples” that do not belong to either class/cluster of interest, but belong to the same domain as the problem at hand. Universum data samples can be obtained easily and are more practical and inexpensive than class labels. We address this topic in feature extraction research and propose a novel semi-supervised approach for feature extraction based on Universum. Experiments are carried out to compare the proposed algorithms with well-known unsupervised and supervised feature extraction algorithms on several UCI data sets. The results show that, with very few Universum data, the proposed algorithms are superior to unsupervised algorithms, and achieve similar or even higher performance than LDA with full class labels on the whole training data.

Community detection is of great importance which enables us to understand the network structure and promotes many real-world applications such as recommendation systems. The heterogeneous social networks, which contain multiple social relations and various user generated content, make the community detection problem more complicated. Particularly, social relations and user generated content are regarded as link information and content information, respectively. Since the two types of information indicate a common community structure from different perspectives, it is better to mine them jointly to improve the detection accuracy. Some detection algorithms utilizing both link and content information have been developed. However, most works take the private community structure of a single data source as the common one, and some methods take extra time transforming the content data into link data compared with mining directly. In this paper, we propose a framework based on regularized joint nonnegative matrix factorization (RJNMF) to utilize link and content information jointly to enhance the community detection accuracy. In the framework, we develop joint NMF to analyze link and content information simultaneously and introduce regularization to obtain the common community structure directly. Experimental results on real-world datasets show the effectiveness of our method.

Nowadays, time series prediction is a hot issue in machine learning, however, how to predict time series fast and accurately remains extremely challengeable. In this paper, we proposed an Improved Error Minimized Extreme Learning Machine (IEM-ELM) algorithm which has better accuracy and prediction on change in direction (POCID) compared with Error Minimized Extreme Learning Machine (EM-ELM) for stock price prediction, meanwhile we implement the Online Sequential algorithm based on IEM-ELM (OSIEM-ELM) which fully inherits the merits of IEM-ELM. The performance of IEM-ELM and OSIEM-ELM are evaluated and compared with EM-ELM and OSEM-ELM respectively, and the experiments are carried out on three stock datasets, experimental results show that IEM-ELM and OSIEM-ELM produces better POCID performance than EM-ELM and OSEM-ELM at fast learning speed.

The Universum data, defined as a set of unlabeled examples that do not belong to any class of interest, have been shown to encode some prior knowledge by representing meaningful information in the same domain as the problem at hand. Universum data have been proved effective in improving learning performance in many tasks, such as classification and clustering. Inspired by its favorable performance, we address a novel semisupervised feature selection problem in this paper, called semisupervised feature selection with Universum, that can simultaneously exploit the unlabeled data and the Universum data. The experiments on several UCI data sets are presented to show that the proposed algorithms can achieve superior performances over conventional unsupervised and supervised methods.

Link prediction is an important task in complex network analysis. Traditional link prediction methods are limited by network topology and lack of node property information, which makes predicting links challenging. In this study, we address link prediction using a sparse Gaussian graphical model and demonstrate its theoretical and practical effectiveness. In theory, link prediction is executed by estimating the inverse covariance matrix of samples to overcome information limits. The proposed method was evaluated with four small and four large real-world datasets. The experimental results show that the area under the curve (AUC) value obtained by the proposed method improved by an average of 3% and 12.5% compared to 13 mainstream similarity methods, respectively. This method outperforms the baseline method, and the prediction accuracy is superior to mainstream methods when using only 80% of the training set. The method also provides significantly higher AUC values when using only 60% in Dolphin and Taro datasets. Furthermore, the error rate of the proposed method demonstrates superior performance with all datasets compared to mainstream methods.

Identification and classification of graph data is a hot research issue in pattern recognition. The conventional methods of graph classification usually convert the graph data to vector representation which ignore the sparsity of graph data. In this paper, we propose a new graph classification algorithm called graph classification based on sparse graph feature selection and extreme learning machine. The key of our method is using lasso to select sparse feature because of the sparsity of the corresponding feature space of the graph data, and extreme learning machine (ELM) is introduced to the following classification task due to its good performance. Extensive experimental results on a series of benchmark graph datasets validate the effectiveness of the proposed methods.

In this work, a new encrypted traffic identification algorithm using sparse logistical regression and extreme learning machine (ELM) is introduced. The proposed method is based on randomness characteristics of encrypted traffic. we utilize l1-norm regularized logistic regression to select sparse features. The identification is performed with the help of Extreme Learning Machine (ELM) because of its better identification and faster speed. In ELM, the input weights and the bias values are randomly chosen and the output weights are analytically calculated. Extensive experiments are performed using the proposed encrypted traffic identification algorithm and results are compared against state of the art techniques.

Extreme learning machine (ELM) has achieved wide attention due to faster learning speed compared with conventional neural network models like support vector machine (SVM) and back-propagation (BP) networks. However, like many other methods, ELM is originally proposed to handle vector pattern while nonvector patterns in real applications need to be explored, such as image data. We propose the two-dimensional extreme learning machine (2DELM) based on the very natural idea to deal with matrix data directly. Unlike original ELM which handles vectors, 2DELM take the matrices as input features without vectorization. Empirical studies on several real image datasets show the efficiency and effectiveness of the algorithm.

Recent studies show that extreme learning machine (ELM) is a suitable, effective, and less time-consuming classifier with a wide range of applications. This chapter addresses the application of ELM to the remotely sensed hyperspectral image classification. In this chapter, the proposed hyperspectral image classification method consists of three steps: First, a semi-supervised feature extract algorithm is used for dimensionality reduction; Second, ELM is taken as a classifier; Finally, conditional random field (CRF) is taken to smooth the result of ELM classifier, where the probability estimation over each class obtained by ELM is used as unary potential function of CRF. The experimental results show that the proposed hyperspectral image classification method using both ELM and CRF achieves good classification performance on two real hyperspectral data sets in comparison to the methods using SVM and CRF.

Data encryption has been widely used. It is important to detect encrypted data. We present a method for detection of encrypted data based on the Support Vector Data Description (SVDD) algorithm. The SVDD is a single class, non-parametric approach for modeling the support of a distribution. We apply the SVDD techniques for detection of encrypted data. Experimental results show that the SVDD can be adopted as an effective tool for detection of encrypted data.

Recently, semi-supervised learning has attracted much attention, and it is applicable to both clustering and classification. But among a large number of these algorithms, only a few considered combining semi-supervised clustering and classification together. However, in supervised learning, the fuzzy relational classifier (FRC) is a recently proposed two-step nonlinear classifier which combines the unsupervised clustering and supervised classification together. Inspired from FRC, in this paper, we present a new method, called Semi-supervised Fuzzy Relational Classifier (SSFRC), which combines semi-supervised clustering and classification together. In the proposed SSFRC, we employ the semi-supervised pair wise-constrained competitive agglomeration (PCCA) to replace FCM to obtain clusters fitting user expectations without specifying the exact cluster number. In addition, we incorporate the fuzzy class labels of unlabeled data into the classification mechanism to improve its performance. The experimental results on real-life datasets demonstrate that SSFRC can outperform FRC with all data labeled in classification performance.

Fuzzy relational classifier(FRC) is a recently proposed nonlinear classifier, in which the unsupervised clustering is performed to explore the underlying structure of the data distribution, and to construct the subsequent classifier. Main advantage of FRC is the interpretable results of the classification prediction. However, the unsupervised FCM is highly sensitive to non-spherical data distribution and improper cluster numbers. In this paper, a new clustering method based on local discriminative information is proposed to group data based on both similarity and class labels, giving rise to that the constructed fuzzy relationship between the formed groups and the given classes is more reliable than unsupervised clustering method. During the classification period, neighbourhood information is incorporated into the classification mechanism to improve its performance. The experimental results on Routers-21578 text dataset and Fudan Chinese textset demonstrate that this new approach has overcome the above disadvantages of FRC and achieved prior robustness and classification performance in cost of small computational cost.

The ability of programming is basic for students to get comprehensive information literacy, so we analysise the problems about programming courses, and put forward that it is important and necessary to reform programming courses teaching model and methods. Through the course of “Data Structures” teaching practice, we study and discuss the programming courses reform content.

Support Vector Data Description (SVDD) as a one-class classifier was developed to construct the minimum hypersphere that encloses all the data of the target class in a high dimensional feature space However, SVDD treats the features of all data equivalently in constructing the minimum hypersphere since it adopts Euclidean distance metric and lacks the incorporation of prior knowledge In this paper, we propose an improved SVDD through introducing relevant metric learning The presented method named RSVDD here assigns large weights to the relevant features and tights the similar data through incorporating the positive equivalence information in a natural way In practice, we introduce relevant metric learning into the original SVDD model with the covariance matrices of the positive equivalence data The experimental results on both synthetic and real data sets show that the proposed method can bring more accurate description for all the tested target cases than the conventional SVDD.

Recently, automatic document classification has attracted a lot of attentions due to the large quantity of web documents. Amongst, a special case is to distinguish whether a document belongs to a target class (Directory) when only the documents of target class are given, which is a standard oneclass classification problem. Moreover, differed from other data, web pages have intrinsic (text) and extrinsic(hyperlink) features. Thus they are very suitable for multiview learning. To tackle the task of one-class document classification, a multiview one-class classifier isproposed, it utilizes the One-cluster clustering based data description (OCCDD) as the base one-class classifier, then gets a one-class classifier in each view by setting a membership threshold, simultaneously, achieves the consensus of different views by a regularization term.Hereafter, different views boost each other, rather than ensemble the results independently orperform document recognition in single view case. We conduct the experiments on the standard WebKBdataset with OCCDD and the proposed multiview method. Experimental results show the good performance of the multiview method in terms of effectiveness and stability to parameter.

As computer networks continue to grow in size and complexity, fault management in today’s high speed telecommunications networks is becoming ever more difficult. As a kernel aspect of network fault management, fault diagnosis by performance data is a process of deducing the exact source of a failure from a set of performance data and fault symptoms. In this paper some existing approach for network fault diagnosis are firstly discussed. It concentrates on analyzing the alarm propagation in the Wide Area Network and a model of fault diagnosis is then proposed. The model is composed of two parts: Self-Organizing Maps training by historical performance data and online fault diagnosis. From our two simulation experimental results with network performance data, our model achieves 96.64 percent detection rate for four kinds of fault types. The performance analysis carried out shows SOM to be a fast and efficient method for fault diagnosis in WAN.

According to the Cost-Sensitive Learning Method, two improved One-Class Anomaly Detection Models using Support Vector Data Description (SVDD) are put forward in this paper. Improved Algorithm is included in the Frequency-Based SVDD (F-SVDD) Model while Input data division method is used in the Write-Related SVDD (W-SVDD) Model. Experimental results show that both of the two new models have a low false positive rate compared with the traditional one. The true positives increased by 22% and 23% while the False Positives decreased by 58% and 94%, which reaches nearly 100% and 0% respectively. And hence, adjusting some parameters can make the false positive rate better. So using Cost-Sensitive method in One-Class Problems may be a future orientation in Trusted Computing area.

Intrusion detection technology is an effective approach to dealing with the problems of network security. In this paper, it presents an intrusion detection model based on neural network and expert system. The key idea is to aim at taking advantage of classification abilities of neural network for unknown attacks and the expert-based system for the known attacks. We employ data from the third international knowledge discovery and data mining tools competition (KDDcup'99) to train and test the feasibility of our proposed neural network component. According to the results of our experiment, our model achieves 96.6 percent detection rate for DOS and probing intrusions, and less than 0.04 percent false alarm rate. Expert system can detect R2L and U2R intrusions more accurately than neural network. Therefore, hybrid model improves the performance to detect intrusions

Intrusion detection technology is an effective approach to dealing with the problems of network security. In this paper, we present an intrusion detection model based on hybrid neural network and C4.5. The key idea is to take advantage of different classification abilities of neural network and the C4.5 algorithm for different attacks. What is more, the model could also be updated by the C4.5 rules mined from the dataset after the event (intrusion). We employ data from the third international knowledge discovery and data mining tools competition (KDDcup '99) to train and test feasibility of our proposed model. From our experimental results with different network data, our model achieves more than 85 percent detection rate on average, and less than 19.7 percent false alarm rate for five typical types of attacks. Through the analysis after-the-event module, the average detection rate of 93.28 percent and false positive rate of 0.2 percent can respectively be obtained.