Search

The figure presents the pre-trained CTR model, the fine-tuned CTR model, and our proposed AutoFT framework from left to right. The pre-trained CTR model with random parameters starting point is optimized based on the data from source domains or based on the data from all domains. The fine-tuned CTR model is initialized with trained parameters in source scenarios and optimized with data from the new target domain. AutoFT involves three sets of parameters: the light-blue boxes contain parameters from the pre-trained CTR model, the dark-blue boxes contain parameters initialized by traditional fine-tuning strategy will be re-optimized during training and the red boxes contain the parameters of the policy networks. [Display omitted] • The proposed AutoFT automatically finds a route between the pre-trained network and the siamese fine-tuned network per instance in the target domain and is compatible with any deep CTR models. • The transfer policy for the embedding layer and the feature interaction layers can be trained synchronously with deep CTR models. • The result shows that the lower layers may represent more general features while higher layers need more fine-tuning to fit a specific target domain. In real business platforms, recommendation systems usually need to predict the CTR of multiple business. Since different scenarios may have common feature interactions, knowledge transferring based methods are often used by re-optimizing the pre-trained CTR model from source scenarios to a new target domain. In addition to knowledge transfer, it is noteworthy that generalizing target domain data outside of the CTR model accurately is also important when re-training all of the fine-tuned parameters. Generally, the pre-trained model trained on large source domains can represent the characteristics of different instances and capture typical feature interactions. It would be useful to directly reuse fine-tuned parameters from source domains to serve the target domain. However, different instances of the target domain may need different amounts of source information to fine-tune the model parameters, and these decisions of freezing or re-optimizing model parameters, which highly depend on the fine-tuned model and target instances, may require much manual effort. In this paper, we propose an end-to-end transfer learning framework with fine-tuned parameters for CTR prediction, called Automatic Fine-Tuning (AutoFT). The principal component of AutoFT is a set of learnable transfer policies that independently determine how the specific instance-based fine-tuning policies should be trained, which decide the routing in the embedding representations and the high-order feature representations layer by layer in deep CTR model. Extensive tests on two benchmarks and one real commercial recommender system deployed in Huawei's App Store show that AutoFT can greatly increase CTR prediction performance when compared to current transferring methodologies. [ABSTRACT FROM AUTHOR]

• Proposed a Novel learn to rank model with explainable results for question routing. • Provided a systematic classification of and the introduction of 74 features that can be effective for the task of question routing. • Offered in-depth analysis on the impact of provided features and feature categories for question routing. • Adopted and reported on the Gini score to enhance interpretability of the results. • Compare the performance of the proposed model against strong neural methods for question routing. Community question answering (CQA) platforms are receiving increased attention and are becoming an indispensable source of information in different domains ranging from board games to physics. The success of these platforms dependent on how efficiently new questions are assigned to community experts, known ascalled question routing. In this paper, we address the problem of question routing by adopting a learning to rank approach over five CQA websites in the context of which we introduce 74 features and systematically classify them into content-based and social-based categories. Our extensive experiments on datasets from five real online question answering websites indicate that content-based features related to tags and topics as well as social features that are related to user characteristics and user temporality are effective for question routing. Our work shows the ability to improve performance compared to the state-of-the-art neural matchmaking methods that lack the interpretability offered by our work. The improvement can be as high as on average 2.47% and 1.10% in terms of common ranking metrics, Normalized Discounted Cumulative Gain (NDCG) and Mean Average Precision (MAP) respectively, compared to our best baselines. [ABSTRACT FROM AUTHOR]

Transfer learning aims to improve the learning of the target domain with the help of knowledge from the source domain. Recently, learning using privileged information (LUPI) has been proposed to learn an accurate classifier with privileged information which is only obtained in the training stage. In this paper, we propose a new AdaBoost-based Transfer Learning with Privileged Information (AdaTLPI) method to solve the transfer learning problem with privileged information, in which AdaBoost is taken into account to combine the weak classifiers into a strong classifier. In the model, we utilize shared parameter to transfer knowledge from the source domain to the target domain. We then incorporate privileged information about the source and target domains into a unified model and AdaBoost is used to learn a strong classifier by combining the obtained weak classifiers. Finally, we present an effective optimization algorithm to solve the proposed objective model and present the boundary of training error of the proposed method. The experimental results manifest that the proposed method can outperform the previous methods. [ABSTRACT FROM AUTHOR]

Knowledge transfer across several streaming processes remain challenging problem not only because of different distributions of each stream but also because of rapidly changing and never-ending environments of data streams. Albeit growing research achievements in this area, most of existing works are developed for a single source domain which limits its resilience to exploit multi-source domains being beneficial to recover from concept drifts quickly and to avoid the negative transfer problem. An online domain adaptation technique under multi-source streaming processes, namely automatic online multi-source domain adaptation (AOMSDA), is proposed in this paper. The online domain adaptation strategy of AOMSDA is formulated under a coupled generative and discriminative approach of denoising autoencoder (DAE) where the central moment discrepancy (CMD)-based regularizer is integrated to handle the existence of multi-source domains thereby taking advantage of complementary information sources. The asynchronous concept drifts taking place at different time periods are addressed by a self-organizing structure and a node re-weighting strategy. Our numerical study demonstrates that AOMSDA is capable of outperforming its counterparts in 5 of 8 study cases while the ablation study depicts the advantage of each learning component. In addition, AOMSDA is general for any number of source streams. The source code of AOMSDA is shared publicly in https://github.com/Renchunzi-Xie/AOMSDA.git. [ABSTRACT FROM AUTHOR]

In recent years, many efficient evolutionary multitasking (EMT) algorithms have been proposed to solve multi-objective multi-task optimization problems. However, EMT algorithms often face negative transfer problems. In this paper, a novel multi-objective evolution strategy, called CT-EMT-MOES, is proposed based on a cultural transmission theory for solving multi-objective multitask optimization problems. First, two evolutionary operators inspired by cultural transmission theory are proposed. The elite-guided variation strategy can transfer the information from the current Pareto front to all individuals and guide the population to quickly converge. The horizontal cultural transmission strategy can efficiently transfer information from the source task to the target task. Second, to solve the negative transfer problem, an adaptive information transfer strategy is proposed to adaptively adjust the probability of an information transfer. Third, the proposed algorithm can gain a Pareto front with good convergence and diversity by utilizing a smaller population size and fewer computing resources. As a result, the proposed algorithm can effectively utilize the implicit similarity and complementarity between simultaneous optimized tasks to improve the overall convergence efficiency and reduce a negative transfer. Finally, comprehensive experimental results show that the proposed algorithm can achieve a better performance compared with previous state-of-the-art multi-objective EMT algorithms. [ABSTRACT FROM AUTHOR]

Multivariate time series classification is a machine learning task with increasing importance due to the proliferation of information sources in different domains (economy, health, energy, crops, etc.). Univariate methods lack the ability to capture the relationships between the different variables that compose a multivariate time series and therefore cannot be directly extrapolated to multivariate environments. Despite the good performance and competitive results of the multivariate proposals published to date, they are hard to interpret due to their high complexity. In this paper, we propose a multivariate time series classification method based on an alternative representation of the time series, composed of a set of 41 descriptive time series features, in order to improve the interpretability of time series and results obtained. Our proposal uses traditional classifiers over the extracted features to look for relationships between the different variables that form a multivariate time series. We have selected four state-of-the-art algorithms as base classifiers to evaluate our method. We have tested our proposal on the complete University of East Anglia repository, obtaining highly interpretable results capable of explaining the relationships between the features that compose the time series and achieving performance results statistically indistinguishable from the best algorithms of the state-of-the-art. [ABSTRACT FROM AUTHOR]

Social networks store a considerable amount of personal data, which are also a source of information for business. To comply with users' privacy rights, all social networks allow users to select the level of privacy they desire. However, what occurs if the privacy choices of a user are modified unilaterally by the social network? The privacy settings chosen by the user are stored by the social network, which acts as a privileged party , which could tamper with the user's choices at any time. This paper addresses this problem and proposes a decentralized approach to manage the privacy settings of a user. Any change in the privacy settings of a social network user is validated by a smart contract to ensure that it is compliant with users' expectations. The proposed solution has been implemented as an Ethereum-based decentralized application to validate the effectiveness of the proposed approach. [ABSTRACT FROM AUTHOR]

This study considers the problem of hybrid community detection in attributed networks based on the information of network topology and attributes with the aim to address the following two shortcomings of existing hybrid community detection methods. First, many of these methods are based on the assumption that network topology and attributes carry consistent information but ignore the intrinsic mismatch correlation between them. Second, network topology is typically treated as the dominant source of information, with attributes employed as the auxiliary source; the dominant effect of attributes is seldom explored or indeed considered. To address these limitations, this paper presents a novel Dual-channel Hybrid Community Detection (DHCD) method that considers the dominant effects of topology and attributes separately. The concept of transition relation between the topology and attribute clusters is introduced to explore the mismatch correlation between the two sources and learn the behavioral and content diversity of nodes. An extended overlapping community detection algorithm is introduced based on the two types of diversity. By utilizing network attributes, DHCD can simultaneously derive the community partitioning membership and corresponding semantic descriptions. The superiority of DHCD over state-of-the-art community detection methods is demonstrated on a set of synthetic and real-world networks. [ABSTRACT FROM AUTHOR]

This study considers the problem of hybrid community detection in attributed networks based on the information of network topology and attributes with the aim to address the following two shortcomings of existing hybrid community detection methods. First, many of these methods are based on the assumption that network topology and attributes carry consistent information but ignore the intrinsic mismatch correlation between them. Second, network topology is typically treated as the dominant source of information, with attributes employed as the auxiliary source; the dominant effect of attributes is seldom explored or indeed considered. To address these limitations, this paper presents a novel Dual-channel Hybrid Community Detection (DHCD) method that considers the dominant effects of topology and attributes separately. The concept of transition relation between the topology and attribute clusters is introduced to explore the mismatch correlation between the two sources and learn the behavioral and content diversity of nodes. An extended overlapping community detection algorithm is introduced based on the two types of diversity. By utilizing network attributes, DHCD can simultaneously derive the community partitioning membership and corresponding semantic descriptions. The superiority of DHCD over state-of-the-art community detection methods is demonstrated on a set of synthetic and real-world networks. [ABSTRACT FROM AUTHOR]

Conversational emotion recognition represents an important machine-learning problem with a wide variety of deployment possibilities. The key challenge in this area is how to properly capture the key conversational aspects that facilitate reliable emotion recognition, including utterance semantics, temporal order, informative contextual cues, speaker interactions as well as other relevant factors. In this paper, we present a novel Graph Neural Network approach for conversational emotion recognition at the utterance level. Our method addresses the outlined challenges and represents conversations in the form of graph structures that naturally encode temporal order, speaker dependencies, and even long-distance context. To efficiently capture the semantic content of the conversations, we leverage the zero-shot feature-extraction capabilities of pre-trained large-scale language models and then integrate two key contributions into the graph neural network to ensure competitive recognition results. The first is a novel context filter that establishes meaningful utterance dependencies for the graph construction procedure and removes low-relevance and uninformative utterances from being used as a source of contextual information for the recognition task. The second contribution is a feature-correction procedure that adjusts the information content in the generated feature representations through a gating mechanism to improve their discriminative power and reduce emotion-prediction errors. We conduct extensive experiments on four commonly used conversational datasets, i.e., IEMOCAP, MELD, Dailydialog, and EmoryNLP, to demonstrate the capabilities of the developed graph neural network with context filtering and error-correction capabilities. The results of the experiments point to highly promising performance, especially when compared to state-of-the-art competitors from the literature. [ABSTRACT FROM AUTHOR]

In the era of explosive data growth, data sources and volumes are rapidly increasing. A multi-source data refers to information from multi-sources. However, not every source of information is equally important; some sources are more important and some are essentially worthless. Therefore, it is very meaningful to study how to select the most valuable sources and to efficiently fuse information. Multi-source incomplete interval-valued data (MSIIV-data) is an important kind of multi-source data. This paper proposes a novel method to information fusion in MSIIV-data via conditional information entropy (CIE) and considers its application to attribute reduction based on mutual information entropy. First, the distance between two information values for each incomplete interval-valued data is defined, the neighborhood classes with a tunable parameter are obtained, and the neighborhood granularity structure is established. Then, a source selection method is given via CIE, which is used to fuse MSIIV-data into single-source incomplete interval-valued data (SSIIV-data). Based on the minimization of CIE, this method allows worthy and reliable information sources to be chosen. Moreover, an attribute reduction algorithm (denoted as MMQPSO) for the fused SSIIV-data is proposed by means of combining mutual information entropy and QPSO-algorithm. Finally, experiments are done to validate the effectiveness of the proposed algorithms. The results of experiment and statistical test on 12 datasets show that the proposed algorithms have certain feasibility and advancement than 6 other advanced algorithms. [ABSTRACT FROM AUTHOR]

Abstract: The cloud computing data center has numerous hosts as well as application requests. It needs to allocate resources dynamically according to the varied demands of users. It should not only provide QoS such as short response time and high throughput but also achieve efficient power consumption. This paper first puts forward a reconfiguration framework based on a request prediction, which anticipates the application request volume in advance. To determine the objective of relatively optimal configuration, it can work out the allocation scheme which can improve the resource utilization ratio as well as lower energy consumption. In addition, a concept of Utility Ratio Matrix (URM) is put forward to represent allocations of hosts and Virtual Machines (VMs), and a reconfiguration algorithm based on request prediction is also presented. The algorithm will predict the application requests so as to work out the allocation scheme in advance. The algorithm can separate the reconfiguration computing from the real allocation so that it can avoid a time delay between the reconfiguration result and the varied demands, and can also reduce the energy consumption in data center. The corresponding analysis and experimental results indicate the feasibility of the reconfiguration algorithm in this paper. [Copyright &y& Elsevier]

In many real-life applications, the data collected from different information sources are always located in diverse sites and characterized by different types of attributes. In this paper, we call this type of data as multi-source hybrid data. Existed rough set models only work well with single source data or multi-source data with one type of attributes, which are not suitable for processing multi-source hybrid data. Furthermore, in dynamic data environment, multi-source hybrid data are typically evolving over time, including the addition of objects and attributes and the revision of missing values in each information source. It will result in the update of information granulation and rough approximation structures. To address these issues, we present an extended rough set model, named as multi-source composite rough sets (MCRS), by integrating different types of attributes and fusing multiple composite relations derived from different information sources. Two novel matrix operators are proposed to construct the matrix-based representation of rough approximations. Then, under the variations of objectives, attributes and attribute values in the multi-source hybrid information system, we present incremental mechanisms based on matrices for the maintenance of approximations by utilizing the previously accumulated matrices information and transmitting location information of part elements in each composite relation matrix. In addition, we develop the incremental algorithms for updating composite rough approximations. A series of comparative experiments show that the proposed algorithm outperforms the static algorithm and other extended incremental algorithms in terms of the efficiency of computation and transmission. [ABSTRACT FROM AUTHOR]

Numerous quantitative graph measures have been defined and applied in various disciplines. Such measures may be differentiated according to whether they are information-theoretic or non-information-theoretic. In this paper, we examine an important property of Randić entropy, an information-theoretic measure, and examine some related graph measures based on random roots. In particular, we investigate the degeneracy of these structural graph measures and discuss numerical results. Finally, we draw some conclusions about the measures' applicability to deterministic and non-deterministic networks. [ABSTRACT FROM AUTHOR]

In D-S evidence theory, belief and commonality functions are well-defined and deeply researched. However, these concepts have not been extended to Random Permutation Set (RPS) because of the difficulties in giving a suitable definition of containment relationships of permutations. In this paper, we addressed this issue by defining both ordered and unordered containment relationships. Based on the defined containment relationships, the corresponding belief function and commonality functions for RPS are derived naturally. Besides, the plausibility function for RPS is also proposed following the definition in D-S evidence theory. Furthermore, the transformations between them and permutation mass functions are discussed and implemented by matrix converters. Thanks to the properties of commonality functions, the combination process in RPS, including left and right orthogonal sum, are both implemented by matrix operations. Moreover, based on matrix operations, the reverse transformation of combination, that is, the decomposition of fusion results given one information source is achieved for the first time. Numerical examples are listed to illustrate the calculation process. The proposed matrix operations provide a way to implement and accelerate information transformations, combinations and decompositions. [ABSTRACT FROM AUTHOR]

Given a temporal network that is composed of a stream of events that evolves continuously over time, like social posts on Reddit, how can we spot interesting events like anomalous posts? In general, it is critical to learn useful information from event time, attributes, and network structure, which means that the anomalous post can be detected by over-frequent posts, impolite comments, and incorrect choices of subreddits. In this paper, we propose a new learning objective, named Continuous-time Graph Directed Information Maximization (CGDIM), to learn informative node presentations for temporal networks. The proposed CGDIM is based on the popular mutual information maximization (InfoMax) method to learn node features that are shared across temporal neighborhoods. Specifically, to capture the time causal relations among edges with continuous time, we utilize directed information rather than mutual information as the measure. The designed learning objective is to maximize the directed information with the direction from the inputs of temporal neighborhoods to the node representations. By setting a source variable that only contains useful information on each node, we prove that directed information is more effective than mutual information for preserving information of the source variables. To show the effectiveness of the CGDIM, we train several popular backbone models for temporal network representation with the CGDIM and test the performance on temporal node/edge classification. Experiment results on four real-world temporal network datasets show that the proposed CGDIM can improve the Average-Precision scores of the backbone models by 3%-8%. [ABSTRACT FROM AUTHOR]

As one commonly used dimensionality reduction method, nonnegative matrix factorization (NMF), whose goal is to learn parts-based representations, has been widely studied and applied to various areas. However, classical NMF does not utilize any label information. In this paper, we propose a novel semi-supervised NMF algorithm named the hyperplane-based nonnegative matrix factorization (HNMF). This HNMF constructs a hyperplane for each cluster such that the labelled points are close to the corresponding hyperplane and far away from the other ones. Then, the discriminative abilities of representation vectors are greatly enhanced. Clustering experiments completed on five publicly available databases demonstrate the effectiveness of this proposed HNMF compared to the state-of-the-art methods. [ABSTRACT FROM AUTHOR]

Abstract: Rough set theory has been proposed by Pawlak as a tool for dealing with the vagueness and granularity in information systems. The core concepts of classical rough sets are lower and upper approximations based on equivalence relations. This paper studies arbitrary binary relation based generalized rough sets. In this setting, a binary relation can generate a lower approximation operation and an upper approximation operation, but some of common properties of classical lower and upper approximation operations are no longer satisfied. We investigate conditions for a relation under which these properties hold for the relation based lower and upper approximation operations. This paper also explores the relationships between the lower or the upper approximation operation generated by the intersection of two binary relations and those generated by these two binary relations, respectively. Through these relationships, we prove that two different binary relations will certainly generate two different lower approximation operations and two different upper approximation operations. [Copyright &y& Elsevier]

Abstract: We use information from the Web for performing our daily tasks more and more often. Locating the right resources that help us in doing so is a daunting task, especially with the present rate of growth of the Web as well as the many different kinds of resources available. The tasks of search engines is to assist us in finding those resources that are apt for our given tasks. In this paper we propose to use the notion of quality as a metric for estimating the aptness of online resources for individual searchers. The formal model for quality as presented in this paper is firmly grounded in literature. It is based on the observations that objects (dubbed artefacts in our work) can play different roles (i.e., perform different functions). An artefact can be of high quality in one role but of poor quality in another. Even more, the notion of quality is highly personal. Our quality-computations for estimating the aptness of resources for searches uses the notion of linguistic variables from the field of fuzzy logic. After presenting our model for quality we also show how manipulation of online resources by means of transformations can influence the quality of these resources. [Copyright &y& Elsevier]

Abstract: The paper considers a wireless information system, wherein various pieces of information represented in XML are broadcast via wireless channels, and mobile clients access the broadcast stream using energy-restricted portable devices. In this paper, we propose a wireless XML streaming method designed to provide energy-efficient access to a wireless stream. We construct two hierarchical structures to represent the XML data and their index information, called the XML data tree and XML index tree, respectively. The wireless XML stream is generated by traversing these two structures with some replications. We design three data/index replication strategies (PP, TT, and TP) in the streaming method. We compare the proposed streaming method with a naı¨ve method called the (1, X) method both analytically and experimentally. Also, based on our analysis results, we determine the optimal method of replication. [Copyright &y& Elsevier]

Abstract: This paper expands and consolidates the use of analogies in thermodynamics to explore concepts in the characterization of information systems. The analogy spans the range of information systems to include databases, knowledge bases and model bases. It includes but is not limited to pressure, expressiveness, temperature, tractability, degrees of order, systems of liquid–liquid equilibrium and disjunction in information-systems integration. By taking advantage of the isomorphism that exists between states of matter and states of information, we can understand new ways to characterize and measure information systems. This paper is the fourth in a series describing new aspects of “infodynamics.”. [Copyright &y& Elsevier]

Computer technology and its relative subjects developed at very high speed in recent years, so is geo-information science, including Geographic Information System (GIS), remote sensing (RS) and global position system (GPS). But with the increase of data, many data cannot be used efficiently because of the tremendous amount of data and information and the difficulty of process and transfer through network. So how to develop internet technology to solve these problems becomes a difficult problem for current computer experts and geo-science experts. Fortunately, grid computing provides us the method to solve this problem effectively. Grid computing is a resources sharing model presented by computer experts to solve current network resources imbalance problem. Basing on the application of grid computing on geographical information system (GIS), this paper analyzes the weakness and problems of traditional GIS, and then gives the method to solve these problems with the technology provided by grid computing and web services. After analyzing the characteristic of grid computing this paper expatiates on current application status of grid computing on GIS and the problems it faces, with the technology of middleware, this paper presents the architecture of grid GIS and lists the techniques it needs. In conclusion, this paper concludes that the distributing middleware architecture based on grid geographic markup language (GridGML) and web services technique is a good solution to current problems, this architecture can also solve those problems such as effective resources sharing through internet and advancing international application's efficiency, at last we discuss its implementation process based on LAN. [ABSTRACT FROM AUTHOR]

Privacy preservation has gained importance with the development of tools for personal information retrieval and social information sharing in Web 2.0 environments. This paper addresses the need for a paradigm shift concerning publicity and privacy together in the process of personal information service and privacy preservation. In addition, we also bridge the conflict between interactivity and independency in the process of social sharing and privacy preservation. By analyzing the paradigm and conflict, we identify the criteria of privacy relativity, such as publicity vs. privacy and interactivity vs. independency. Using hierarchical clustering methods, we analyze the hierarchical community characteristics of tag networks which reflect users’ preferences and the features of information resources through social tagging. Based on the hierarchical communities of tag networks and relevant parameters of partitions such as the k of k -clique and the number of communities N at a certain level, we can determine and adjust the degree and the scope of users’ preferences and the features of information resources which can be revealed according to users’ permissions. The adjustment process can reflect the trade-off between privacy preservation and information access. We believe that this work will prove to be an important first-step toward the personalization and socialization of security policies in Web 2.0 environments. [ABSTRACT FROM AUTHOR]

Abstract: The World Wide Web is the most rapidly growing and accessible source of information. Its popularity has been largely influenced by the wide availability of the Internet in almost every modern house and even on the go after the wide-spread of the handheld devices. Yet, pages on the Web have an additional template (we call it noisy) information that does not add value to the actual content of the page. Even worse, it can harm the effectiveness of Web mining techniques; these templates could be eliminated by preprocessing. Templates form one popular type of noise on the Internet. In this paper, we introduce Noise Detector (ND) as an effective approach for detecting and removing templates from Web pages. ND segments Web pages into semantically coherent blocks. Then it computes content and structure similarities between these blocks; a presentational noise measure is used as well. ND dynamically calculates a threshold for differentiating noisy blocks. Provided that the investigated website has a single visible template, ND can detect the template with high accuracy using two pages only. However, ND can be expanded to detect multiple templates per website, and the challenge will be to minimize the number of pages to be checked. Further, ND leads to website summarization. The conducted experiments show that ND outperforms existing approaches in space complexity, time complexity (see Section 4.6 for more details on ND’s processing time against other algorithms’), minimum requirements to produce acceptable results, and results accuracy. [Copyright &y& Elsevier]

Abstract: There are mainly two methodologies dealing with uncertainty measurement issue in rough set theory: pure rough set approach and information theory approach. Pure rough set approach is based on the concepts of accuracy, roughness and approximation accuracy proposed by Pawlak. Information theory approach is based on Shannon’s entropy or its variants. Several authors have extended the information theory approach into incomplete information systems. However, there are few studies on extending the pure rough set approach to incomplete information systems. This paper focuses on constructing uncertainty measures in incomplete information systems by pure rough set approach. Three types of definitions of lower and upper approximations and corresponding uncertainty measurement concepts including accuracy, roughness and approximation accuracy are investigated. Theoretical analysis indicates that two of the three types can be used to evaluate the uncertainty in incomplete information systems. Experiments on incomplete real-life data sets have been conducted to test the two selected types (the first type and the third type) of uncertainty measures. Results show that the two types of uncertainty measures are effective. [Copyright &y& Elsevier]

Abstract: In a world with vast information overload, well-optimized retrieval of relevant information has become increasingly important. Dividing large, multiple topic spanning documents into sets of coherent subdocuments facilitates the information retrieval process. This paper presents a novel technique to automatically subdivide a textual document into consistent components based on a coherence quantification function. This function is based on stem or term chains linking document entities, such as sentences or paragraphs, based on the reoccurrences of stems or terms. Applying this function on a document results in a coherence graph of the document linking its entities. Spectral graph partitioning techniques are used to divide this coherence graph into a number of subdocuments. A novel technique is introduced to obtain the most suitable number of subdocuments. These subdocuments are an aggregation of (not necessarily adjacent) entities. Performance tests are conducted in test environments based on standardized datasets to prove the algorithm’s capabilities. The relevance of these techniques for information retrieval and text mining is discussed. [Copyright &y& Elsevier]

Abstract: Lin and Tasi, Yang et al., and Chang et al.’s meaningful secret sharing schemes provided authentication mechanisms but none included a remedy ability that would cause the secret image never to be completely obtained while some information of the stego-images are losing or tampering with. This paper proposes a meaningful secret-sharing scheme which includes both authentication and remedy abilities that allow for detection of the corrupted area and use of the hidden information to repair the secret image with reasonable visual quality. In comparison with previous schemes, this approach results in superior visual qualities of the stego-images by an average of more than 3dB. [Copyright &y& Elsevier]

Abstract: Web-based databases are gaining increased popularity. This has positively influenced the availability of structured and semi-structured databases for access by a variety of users ranging from professionals to naive users. The number of users accessing online databases will continue to increase if the visual tools connected to web-based databases are flexible and user-friendly enough to meet the expectations of naive users and professionals. Further, XML is accepted as the standard for platform independent data exchange. This motivated for the development of the conversion tools between structured databases and XML. Realizing that such a need has not been well handled by the available tools, including Clio from IBM, we developed VIREX as a visual tool for converting relational databases into XML, and since then has been empowered with further capabilities to manipulate the produced XML schema including the maintenance of materialized views and schema evolution functions. VIREX provides an interactive approach for querying and integrating relational databases to produce XML documents and the corresponding XML schema(s). VIREX supports VRXQuery as a visual naive users-oriented query language that allows users to specify queries and define views directly on the interactive diagram as a sequence of mouse clicks with minimum keyboard input. As the query result, VIREX displays on the screen the XML schema that satisfies the specified characteristics and generates colored (easy to read) XML document(s). The main contribution described in this paper is the novel approach for turning query results into materialized views which are maintained to remain consistent with the underlying database. VIREX supports deferred update of XML views by keeping an ordered summary of the necessary and sufficient information required for the process. Each view has a corresponding marker in the ordered summary to indicate the start of the information to be reflected onto the view when it is accessed. When a view is accessed, its marker moves to the head of the list to mark for the next update. In addition, VIREX supports some basic schema evolution functions include renaming, adding and dropping of elements and attributes, among others. The supported schema evolution functions add flexibility to the view maintenance and materialization process. [Copyright &y& Elsevier]

Abstract: Nowadays, various promising paradigms of distributed computing over the Internet, such as Grids, P2P and Clouds, have emerged for resource sharing and collaboration. To enable resources sharing and collaboration across different domains in an open computing environment, virtual organizations (VOs) often need to be established dynamically. However, the dynamic and autonomous characteristics of participating domains pose great challenges to the security of virtual organizations. In this paper, we propose a secure collaboration service, called PEACE-VO, for dynamic virtual organizations management. The federation approach based on role mapping has extensively been used to build virtual organizations over multiple domains. However, there is a serious issue of potential policy conflicts with this approach, which brings a security threat to the participating domains. To address this issue, we first depict concepts of implicit conflicts and explicit conflicts that may exist in virtual organization collaboration policies. Then, we propose a fully distributed algorithm to detect potential policy conflicts. With this algorithm participating domains do not have to disclose their full local privacy policies, and is able to withhold malicious internal attacks. Finally, we present the system architecture of PEACE-VO and design two protocols for VO management and authorization. PEACE-VO services and protocols have successfully been implemented in the CROWN test bed. Comprehensive experimental study demonstrates that our approach is scalable and efficient. [Copyright &y& Elsevier]

Abstract: The total number of popular search engines has decreased over time from its peak of the late 1990s. However, when combining the three remaining major ones (Yahoo!, Google, MS Live) with large repositories of information (e.g. BBC.com, NYT.com, Gazeta.pl, etc.), the total number of important information sources can be seen as slowly increasing. Focusing on utilization of search engines only, it is easy to observe that the same query issued to each one of them results in different “suggestions.” The question thus arrives, is it possible (and worthy) to combine responses obtained from each one of them into a single answer set. In this paper, we look into three approaches of achieving this goal, which are based on: Game theory, Auction and Consensus methods, while our focus is to study (and compare) their “performance.” [Copyright &y& Elsevier]

Abstract: In this paper, we present a method based on fuzzy transforms to establish dependencies between numerical attributes in datasets. We find the best fuzzy partitions of the attribute domains with respect to which we determine the direct and inverse fuzzy transforms. We use two specific regression indexes (which must be smaller than a threshold deduced experimentally) for evaluating dependency between numerical attributes. The experiments are conducted on two well known datasets: “El Nino” (http://kdd.ics.uci.edu/databases/el_nino/el_nino.data.html) and the remote sensing data determined from US Forest Service (Region 2, Resource Information System, http://kdd.ics.uci.edu/databases/covertype/covertype.data.html). Our results are quite in agreement with the regression analysis of the same data. [Copyright &y& Elsevier]

Abstract: Stacking is a general ensemble method in which a number of base classifiers are combined using one meta-classifier which learns their outputs. Such an approach provides certain advantages: simplicity; performance that is similar to the best classifier; and the capability of combining classifiers induced by different inducers. The disadvantage of stacking is that on multiclass problems, stacking seems to perform worse than other meta-learning approaches. In this paper we present Troika, a new stacking method for improving ensemble classifiers. The new scheme is built from three layers of combining classifiers. The new method was tested on various datasets and the results indicate the superiority of the proposed method to other legacy ensemble schemes, Stacking and StackingC, especially when the classification task consists of more than two classes. [Copyright &y& Elsevier]

Abstract: In this paper, we propose two risk-sensitive loss functions to solve the multi-category classification problems where the number of training samples is small and/or there is a high imbalance in the number of samples per class. Such problems are common in the bio-informatics/medical diagnosis areas. The most commonly used loss functions in the literature do not perform well in these problems as they minimize only the approximation error and neglect the estimation error due to imbalance in the training set. The proposed risk-sensitive loss functions minimize both the approximation and estimation error. We present an error analysis for the risk-sensitive loss functions along with other well known loss functions. Using a neural architecture, classifiers incorporating these risk-sensitive loss functions have been developed and their performance evaluated for two real world multi-class classification problems, viz., a satellite image classification problem and a micro-array gene expression based cancer classification problem. To study the effectiveness of the proposed loss functions, we have deliberately imbalanced the training samples in the satellite image problem and compared the performance of our neural classifiers with those developed using other well-known loss functions. The results indicate the superior performance of the neural classifier using the proposed loss functions both in terms of the overall and per class classification accuracy. Performance comparisons have also been carried out on a number of benchmark problems where the data is normal i.e., not sparse or imbalanced. Results indicate similar or better performance of the proposed loss functions compared to the well-known loss functions. [Copyright &y& Elsevier]

Abstract: Knowledge encoded in information systems can be represented by different sets of rules generated by these systems. One can consider sets of deterministic, nondeterministic or probabilistic rules. Such sets of rules can be treated as theories of information systems. Any such a theory generated from a given information system corresponds to a subjective view on knowledge encoded in this information system. Such theories can be used for solving different problems. For example, the maximal consistent extensions of information systems were studied for synthesis of concurrent processes specified by information systems. In this approach, the maximal consistent extension of a given information system consists of all objects perceived by means of attributes which are consistent with the theory including all the so called true and realizable deterministic rules extracted from the original information system. In this paper, we report results on the maximal consistent extensions of information systems relative to some other theories of information systems, e.g., theories consisting of rules such as true and realizable inhibitory rules, true inhibitory rules, and true deterministic rules. We also discuss algorithmic problems related to the maximal consistent extensions. In particular, from the obtained results it follows that solutions based on these new sets of rules, e.g., on inhibitory rules can be of higher quality than in the case of deterministic rules. [Copyright &y& Elsevier]

Abstract: Attribute reduction is a basic issue in knowledge representation and data mining. This paper deals with attribute reduction in incomplete information systems and incomplete decision systems based on Dempster–Shafer theory of evidence. The concepts of plausibility reduct and belief reduct in incomplete information systems as well as relative plausibility reduct and relative belief reduct in incomplete decision systems are introduced. It is shown that in an incomplete information system an attribute set is a belief reduct if and only if it is a classical reduct and a plausibility consistent set must be a classical consistent set. In a consistent incomplete decision system, the concepts of relative reduct, relative plausibility reduct, and relative belief reduct are all equivalent. In an inconsistent incomplete decision system, an attribute set is a relative plausibility reduct if and only if it is a relative reduct, a plausibility consistent set must be a belief consistent set, and a belief consistent set is not a plausibility consistent set in general. [Copyright &y& Elsevier]

Abstract: Many methods based on the rough set to deal with incomplete information systems have been proposed in recent years. However, they are only suitable for the incomplete systems with regular attributes whose domains are not preference-ordered. This paper thus attempts to present research focusing on a complex incomplete information system—the incomplete ordered information system. In such incomplete information systems, all attributes are considered as criterions. A criterion indicates an attribute with preference-ordered domain. To conduct classification analysis in the incomplete ordered information system, the concept of similarity dominance relation is first proposed. Two types of knowledge reductions are then formed for preserving two different notions of similarity dominance relations. With introduction of the approximate distribution reduct into the incomplete ordered decision system, the judgment theorems and discernibility matrixes associated with four novel approximate distribution reducts are obtained. A numerical example is employed to substantiate the conceptual arguments. [Copyright &y& Elsevier]

Abstract: One of the key problems of knowledge discovery is knowledge reduction. This paper proposes a new method for knowledge reduction in information systems. First, two families of closed sets and are defined, where r and R are equivalence relations defined on the attribute set and its power set, respectively. The properties of and are also discussed. The necessary and sufficient condition for is then given and employed to construct an approach to attribute reduction in information systems. It is also proved that under the condition , the proposed approach to knowledge reduction is equivalent to the well-accepted one in reference [W.X. Zhang, Y. Leung, W.Z. Wu, Information Systems and Knowledge Discovery, Science Publishing Company, Beijing, 2003]. [Copyright &y& Elsevier]

Abstract: This paper proposes an efficient pattern extraction algorithm that can be applied on melodic sequences that are represented as strings of abstract intervallic symbols; the melodic representation introduces special “binary don’t care” symbols for intervals that may belong to two partially overlapping intervallic categories. As a special case the well established “step–leap” representation is examined. In the step–leap representation, each melodic diatonic interval is classified as a step (±s), a leap (±l) or a unison (u). Binary don’t care symbols are used to represent the possible overlapping between the various abstract categories e.g. , and , . We propose an -time algorithm for computing all maximal-pairs in a given sequence , where x contains d occurrences of binary don’t cares and z is the number of reported maximal-pairs. [Copyright &y& Elsevier]

Abstract: The mobile Internet provides mobile nodes with continuous access to the Internet resources. This type of service is guaranteed only if these mobile nodes are one-hop away from the edge of the Internet. Mobile ad hoc networks (MANET), however, enable multi-hop communication between mobile nodes. A more promising goal would be to connect MANET to the Internet, thus enabling mobile nodes to benefit from the multi-hop nature of MANET so they can to share access to the Internet resources with the stationary nodes. This paper aims to study a three-tier architecture to provide MANET nodes with multi-hop Internet connectivity, evaluate its performance through extensive experiments with respect to a set of decision parameters, and outline the lessons learned from these experiments. [Copyright &y& Elsevier]

Abstract: Words mean different things to different people, and so are uncertain. We, therefore, need a fuzzy set model for a word that has the potential to capture their uncertainties. In this paper I propose that an interval type-2 fuzzy set (IT2 FS) be used as a FS model of a word, because it is characterized by its footprint of uncertainty (FOU), and therefore has the potential to capture word uncertainties. Two approaches are presented for collecting data about a word from a group of subjects and then mapping that data into a FOU for that word. The person MF approach, in which each person provides their FOU for a word, is limited to fuzzy set experts because it requires the subject to be knowledgeable about fuzzy sets. The interval end-points approach, in which each person provides the end-points for an interval that they associate with a word on a prescribed scale is not limited to fuzzy set experts. Both approaches map data collected from subjects into a parsimonious parametric model of a FOU, and illustrate the combining of fuzzy sets and statistics—type-2 fuzzistics. [Copyright &y& Elsevier]

Abstract: Information systems, which contain only crisp data, precise and unique attribute values for all objects, have been widely investigated. Due to the fact that in realworld applications imprecise data are abundant, uncertainty is inherent in real information systems. In this paper, information systems are called fuzzy information systems, and formalized by (objects; attributes; f), in which f is a fuzzy set and expresses some uncertainty between an object and its attribute values. To interpret and extract fuzzy decision rules from fuzzy information systems, the meta-theory based on modal logic proposed by Resconi et al. is modified. The modified meta-theory not only expresses uncertainty between objects and their attributes, but also uncertainty in the process of recognizing fuzzy information systems. In addition, according to perception computing (proposed by Zadeh), granules of fuzzy information systems can be represented by fuzzy decision rules, so that, fuzzy inference methods can be used to obtain the decision attribute of a new object. Finally, a novel way of combining evidences based on the modified meta-theory is introduced, which extends the concept of combining evidences based on Dempster–Shafer theory. [Copyright &y& Elsevier]

Statistical language models encapsulate varied information, both grammatical and semantic, present in a language. This paper investigates various techniques for overcoming the difficulties in modelling highly inflected languages. The main problem is a large set of different words. We propose to model the grammatical and semantic information of words separately by splitting them into stems and endings. All the information is handled within a data-driven formalism. Grammatical information is well modelled by using short-term dependencies. This article is primarily concerned with the modelling of semantic information diffused through the entire text. It is presumed that the language being modelled is homogeneous in topic. The training corpus, which is very topically heterogeneous, is divided into three semantic levels based on topic similarity with the target environment text. Text on each semantic level is used as training text for one component of a mixture model. A document is defined as a basic unit of a training corpus, which is semantically homogeneous. The similarity of topic between a document and a collection of target environment texts is determined by the cosine vector similarity function and TFIDF weighting heuristic. The crucial question in the case of highly inflected languages is how to define terms. Terms are defined as clusters of words. Clustering is based on approximate string matching. We experimented with Levenshtein distance and Ratcliff/Obershelp similarity measure, both in combination with ending-stripping. Experiments on the Slovenian language were performed on a corpus of VECER newswire text. The results show a significant reduction in OOV rate and perplexity. [ABSTRACT FROM AUTHOR]

Although E-mail systems are one of the most useful communication tools for business, education, etc., missing important E-mail messages become a very serious problem. It is very useful filtering supports for users to pick up important messages or to neglect unnecessary messages. This paper presents a method of determining the time priority for E-mail messages. Multi-attribute rules are defined to detect complex time expressions and a set pattern-matching machine is proposed. It enables us to protect missing messages with important time information because the presented method can classify and rank them according to time priority measurement automatically. From the simulation results of determining time priority, the presented pattern-matching method is about 4 times faster than the traditional string pattern-matching method. From the results of filtering 5172 sentences, it is verified that precision and recall of the presented method becomes 95% and 96%, respectively. From the experimental results of determining 10 highest messages among 100 E-mail, it is verified that filtering time by the proposed measurement is from 9.7 to 16.6 faster than a non-filtering method. [ABSTRACT FROM AUTHOR]

A tolerance is a reflexive and symmetric, but not necessarily transitive, binary relation. Contrary to what happens with equivalence relations, when dealing with tolerances one must distinguish between blocks (maximal subsets where the tolerance is a total relation) and classes (the class of an element is the set of those elements tolerable with it). Both blocks and classes of a tolerance on a set define coverings of this set, but not every covering of a set is defined in this way. The characterization of those coverings that are families of blocks of some tolerance has been known for more than a decade now. In this paper we give a characterization of those coverings of a finite set that are families of classes of some tolerance. [ABSTRACT FROM AUTHOR]

The purpose of this paper is to introduce two new spaces and also to define Stσ and st0σ of strongly σ-statistically convergent and lacunary strongly σ-statistically convergent sequences, respectively. Also, we give some inclusion relations involving these spaces. [ABSTRACT FROM AUTHOR]

This paper proposes a new hybrid case-based architecture, which supports multiple- disease diagnosis and the learning of new adaptation knowledge. The architecture combines case-based reasoning (CBR), neural networks, fuzzy theory, induction, utility theory, and knowledge-based planning technology together to facilitate medical diagnosis. The basic mechanism is that of CBR. A distributed fuzzy neural network is employed to perform approximate matching and thus to tolerate potential noise in case retrieval. The induction technology along with utility theory is used to select valuable features of the target case and prune unnecessary search space. Knowledge-based planning is a general-purpose mechanism for case adaptation. It creates a case adaptation plan from an adaptation tree, which contains all relevant problem features, satisfies all relevant constraints, and contains all cases whose expected utilities are greater than a threshold. Execution of the case adaptation plan leads to the diagnosis of multiple diseases. The adaptation tree facilitates the reuse of cases and the learning of various types of knowledge-including relationships between disease types and features, case-specific verification knowledge, and differential diagnosis rules. Integrating these techniques in the CBR paradigm can effectively produce a high-quality diagnosis for a given medical consultation. [ABSTRACT FROM AUTHOR]

In this paper, two new operators are introduced for the rough set theory. Using them, two inequalities well known in the rough set theory can now be modified to become equalities. With this change, no information will be lost in the new expressions. Hence, many properties in rough set theory can be improved and in particular, the union, the intersection, and the complement operations can be redefined based on the two equalities. Furthermore, the collection of rough sets of an approximation space forms a Boolean algebra under these new operators. Finally, roughness properties of fuzzy sets are analyzed using the new operations. [ABSTRACT FROM AUTHOR]

In this paper a new method is developed to create a space surface interpolation using only values of the function being interpolated. This method has some characteristics comparing with the present interpolation method. First, the interpolation function has a simple and explicit mathematical representation, which is convenient both in practical application and in theoretical study. Secondly, the interpolating surfaces are C' in the interpolating region under the condition that the interpolation is only based on the function values. Third, the interpolation surfaces could be varied as the parameters vary, thus, the interpolating surfaces can be modified by selecting suitable parameters under the condition that the interpolation data are not changed, and more important theoretically is that the uniqueness of the interpolating function for the given data would be replaced by the uniqueness of the interpolation for the given data and the parameters. [ABSTRACT FROM AUTHOR]

Testing and debugging remove faults from software and ensure its reliable operation. In software developments, enormous resources are spent on testing and debugging, and the efficiencies of them are important factors when we consider software development cost. However, testing methods have been evaluated mainly on effectiveness, such as a coverage of source codes, and efficiencies of the methods have not been evaluated and no applicable criterion exists. In this paper, we firstly indicate the criterion of testing effect and efficiencies with search theory, we define the efficiency ratio of testing by duplication of the detected fault, and we define the efficiency ratio of debugging by delay time of repair. Then we evaluate actual efficiency ratio using the criterion and case data. This evaluation method considers testing and debugging as search processes and visualizes the search structure of software. As a result of analysis, it has been indicated that testing and debugging efficiencies is determined by duplication of search path on search structure, and it is possible to make extensive improvements of testing and debugging efficiencies. [ABSTRACT FROM AUTHOR]

This paper presents a new metaheuristic approach called ACOMAC algorithm for solving the traveling salesman problem (TSP). We introduce multiple ant clans' concept from parallel genetic algorithm to search solution space utilizing various islands to avoid local minima and thus can yield global minimum for solving the traveling salesman problem. Moreover, we present two approaches named the multiple nearest neighbor (NN) and the dual nearest neighbor (DNN) to ACOMAC to enhance large TSPs. To validate the proposed methods, numerous simulations were conducted to compare ACOMAC and Dorigo's ACS with and without the addition of the multiple nearest neighbor (NN) method or the dual nearest neighbor (DNN) approach, using a range of TSP benchmark problems. According to the results of the simulation, adding the NN or DNN approach to ACOMAC or ACS, as initial solutions, also significantly enhances the performance of ACO MAC and ACS in solving the traveling salesman problem. Meanwhile, using ACOMAC + DNN with TSP can yield better solutions than the other stated approaches. Additionally, ACOMAC or ACOMAC + NN, utilizing five ant clans with a total of 20 ants, is verified to yield better solutions. Furthermore, ACOMAC with a local weighting (w) set to 0.6 can yield better solutions in terms of length. [ABSTRACT FROM AUTHOR]