With the advance of internet, multi-modal data can be easily collected from many social websites such as Wikipedia, Flickr, YouTube, etc. Images shared on the web are usually associated with social tags or other textual information. Although existing multi-modal methods can make use of associated text to improve image annotation, the disadvantages of them are that associated text is also required for a new image to be predicted. In this paper, we propose the cross-modal multi-task learning (CMMTL) framework for image annotation. Labeled and unlabeled multi-modal data are both levaraged for training in CMMTL, and it finally obtains visual classifiers which can predict concepts for a single image without any associated information. CMMTL integrates graph learning, multi-task learning and cross-modal learning into a joint framework, where a shared subspace is learned to preserve both cross-modal correlation and concept correlation. The optimal solution of the proposed framework can be obtained by solving a generalized eigenvalue problem. We conduct comprehensive experiments on two real world image datasets: MIR Flickr and NUS-WIDE, to evaluate the performance of the proposed framework. Experimental results demonstrate that CMMTL obtains a significant improvement over several representative methods for cross-modal image annotation.

A large number of practical domains, such as scene classification and object recognition, have involved more than two classes. Therefore, how to directly conduct multiclass classification is being an important problem. Although some multiclass boosting methods have been proposed to deal with the problem, the combinations of weak learners are confined to linear operation, namely weighted sum. In this paper, we present a novel large-margin loss function to directly design multiclass classifier. The resulting risk, which guarantees Bayes consistency and global optimization, is minimized by gradient descent or Newton method in a multidimensional functional space. At every iteration, the proposed boosting algorithm adds the best weak learner to the current ensemble according to the corresponding operation that can be sum or Hadamard product. This process grown in an adaptive manner can create the sum of Hadamard products of weak learners, leading to a sophisticated nonlinear combination. Extensive experiments on a number of UCI datasets show that the performance of our method consistently outperforms those of previous multiclass boosting approaches for classification.

Regression test case reduction aims at selecting a representative subset from the original test pool, while retaining the largest possible fault detection capability. Cluster analysis has been proposed and applied for selecting an effective test case subset in regression testing. It groups test cases into clusters based on the similarity of historical execution profiles. In previous studies, historical execution profiles are represented as binary or numeric function coverage vectors. The vector-based similarity approaches only consider which functions or statements are covered and the number of times they are executed. However, the vector-based approaches do not take the relations and sequential information between function calls into account. In this paper, we propose cluster analysis of function call sequences to attempt to improve the fault detection effectiveness of regression testing even further. A test is represented as a function call sequence that includes the relations and sequential information between function calls. The distance between function call sequences is measured not only by the Levenshtein distance but also the Euclidean distance. To assess the effectiveness of our approaches, we designed and conducted experimental studies on five subject programs. The experimental results indicate that our approaches are statistically superior to the approaches based on the similarity of vectors (i.e. binary vectors and numeric vectors), random and greedy function-coverage-based maximization test case reduction techniques in terms of fault detection effectiveness. With respective to the cost-effectiveness, cluster analysis of sequences measured using the Euclidean distance is more effective than using the Levenshtein distance.

Combinatorial interaction testing is a well-studied testing strategy, and has been widely applied in practice. Combinatorial interaction test suite, such as fixed-strength and variable-strength interaction test suite, is widely used for combinatorial interaction testing. Due to constrained testing resources in some applications, for example in combinatorial interaction regression testing, prioritization of combinatorial interaction test suite has been proposed to improve the efficiency of testing. However, nearly all prioritization techniques may only support fixed-strength interaction test suite rather than variable-strength interaction test suite. In this paper, we propose two heuristic methods in order to prioritize variable-strength interaction test suite by taking advantage of its special characteristics. The experimental results show that our methods are more effective for variable-strength interaction test suite by comparing with the technique of prioritizing combinatorial interaction test suites according to test case generation order, the random test prioritization technique, and the fixed-strength interaction test suite prioritization technique. Besides, our methods have additional advantages compared with the prioritization techniques for fixed-strength interaction test suite.

Automatic image annotation is an important task for multimedia retrieval. By allocating relevant words to un-annotated images, these images can be retrieved in response to textual queries. There are many researches on the problem of image annotation and most of them construct models based on joint probability or posterior probabilities of words. In this paper we estimate the probabilities that words generate the images, and propose a two-phase generation model for the generation procedure. Each word first generates its related words, then these words generate an un-annotated image, and the relation between the words and the un-annotated image is obtained by the probability of the two-phase generation. The textual words usually contain more semantic information than visual content of images, thus the probabilities that words generate images is more reliable than the probability that images generate words. As a result, our model estimates the more reliable probability than other probabilistic methods for image annotation. The other advantage of our model is the relation of words is taken into consideration. The experimental results on Corel 5K and MIR Flickr demonstrate that our model performs better than other previous methods. And two-phase generation which considering word's relation for annotation is better than one-phase generation which only consider the relation between words and images. Moreover, the methods which estimate the generative probability obtain better performance than SVM which estimates the posterior probability.

In this paper, a semantic model for cross-modal and multi-modal retrieval is studied. We assume that the semantic correlation of multimedia data from different modalities can be depicted in a probabilistic generation framework. Media data from different modalities can be generated by the same semantic concepts, and the generation process of each media data is conditional independent under the semantic concepts. The semantic generation model (SGM) for cross-modal and multi-modal analysis is proposed based on this assumption. We study two types of methods: direct method Gaussian distribution and indirect method random forest, to estimate the semantic conditional distribution of SGM. Then methods for cross-modal and multi-modal retrieval are derived from SGM. Experimental results show that SGM based methods for cross-modal retrieval improve the accuracy over the state-of-the-art cross-modal method, but don't increase the time consuming, and the SGM multimodal retrieval methods also outperform traditional methods in image retrieval. Moreover, indirect SGM based method outperforms direct SGM method in the two types of retrieval, which proves that indirect SGM can better describe the semantic distribution.

Combinatorial interaction testing is a well-recognized testing method, and has been widely applied in practice, often with the assumption that all test cases in a combinatorial test suite have the same fault detection capability. However, when testing resources are limited, an alternative assumption may be that some test cases are more likely to reveal failure, thus making the order of executing the test cases critical. To improve testing cost-effectiveness, prioritization of combinatorial test cases is employed. The most popular approach is based on interaction coverage, which prioritizes combinatorial test cases by repeatedly choosing an unexecuted test case that covers the largest number of uncovered parameter value combinations of a given strength (level of interaction among parameters). However, this approach suffers from some drawbacks. Based on previous observations that the majority of faults in practical systems can usually be triggered with parameter interactions of small strengths, we propose a new strategy of prioritizing combinatorial test cases by incrementally adjusting the strength values. Experimental results show that our method performs better than the random prioritization technique and the technique of prioritizing combinatorial test suites according to test case generation order, and has better performance than the interaction-coverage-based test prioritization technique in most cases.

Random testing (RT), a fundamental software testing technique, has been widely used in practice. Adaptive random testing (ART), an enhancement of RT, performs better than original RT in terms of fault detection capability. However, not much work has been done on effectiveness analysis of ART in the combinatorial test spaces. In this paper, we propose a novel family of ART-based algorithms for generating combinatorial test suites, mainly based on fixed-size-candidate-set ART and restricted random testing (that is, ART by exclusion). We use an empirical approach to compare the effectiveness of test sets obtained by our proposed methods and random selection strategy. Experimental data demonstrate that the ART-based tests cover all possible combinations at a given strength more quickly than randomly chosen tests, and often detect more failures earlier and with fewer test cases in simulations.

Unreliable component security hinders the development of component technology. Component security testing is rarely researched with comprehensive focus; several approaches or technologies for detecting vulnerabilities in component security have been proposed, but most are infeasible. A testing approach for component security, which is based on the chemical abstract machine, is proposed for detecting explicit and implicit component security vulnerabilities. We develop an extended chemical abstract machine model, called eCHAM, and generate a state transfer tree and testing sequence for components based on the proposed model. The model can help test the explicit security exceptions of components according to the testing approach of interface fault injection. Condition and state mutation algorithms for identifying implicit security exceptions are also proposed. Vulnerability testing reports are obtained according to the test results. Experiments were conducted in an integration testing platform to verify the applicability of the proposed approach. Results show that the approach is effective and practicable. The proposed approach can detect explicit and implicit security exceptions of components.

Given a data point set D, a query point set Q, and an integer k, the Group Nearest Group (GNG) query finds a subset ω (|ω| ≤ k)of points from Dsuch that the total distance from all points in Q to the nearest point in ω is not greater than any other subset ω' (|ω'| ≤ k) of points in D. GNG query is a partition-based clustering problem which can be found in many real applications and is NP-hard. In this paper, Exhaustive Hierarchical Combination (EHC) algorithm and Subset Hierarchial Refinement (SHR) algorithm are developed for GNG query processing. While EHC is capable to provide the optimal solution for k = 2, SHR is an efficient approximate approach that combines database techniques with local search heuristic. The processing focus of our approaches is on minimizing the access and evaluation of subsets of cardinality k in D since the number of such subsets is exponentially greater than |D|. To do that, the hierarchical blocks of data points at high level are used to find an intermediate solution and then refined by following the guided search direction at low level so as to prune irrelevant subsets. The comprehensive experiments on both real and synthetic data sets demonstrate the superiority of SHR in terms of efficiency and quality.

The goal of this paper is to minimize service disruption, a unique phenomenon in peer-to-peer (P2P) streaming system and the major contributing factor to degrade user experience. In order to improve its streaming quality and stability, a P2P system should select peers with abundant bandwidth resource and prospect longevity as the parents of other peers. Towards this end, we propose PRW (peer resilient weight), a metric to evaluate a peer's healthiness regarding these two aspects. The theoretical background of PRW originates from the optimization framework based on the generalized flow theory. To show PRW's applicability to a wide variety of P2P streaming solutions, we propose several parent selection algorithms guided by PRW under the single-tree, multi-tree, and mesh structures. Using MSN and PPLive traces, we find our algorithms to be able to achieve lower service disruption and lower peer rejection, compared to several existing heuristics and algorithms.

SUMMARY Random testing (RT), which simply selects test cases at random from the whole input domain, has been widely applied to test software and assess the software reliability. However, it is controversial whether RT is an effective method to detect software failures. Adaptive random testing (ART) is an enhancement of RT in terms of failure‐detection effectiveness. Its basic intuition is to evenly spread random test cases all over the input domain. There are various notions to achieve the goal of even spread, and each notion can be implemented by different algorithms. For example, ‘by exclusion’ and ‘by partitioning’ are two different notions to evenly spread test cases. Restricted random testing (RRT) is a typical algorithm for the notion of ‘by exclusion’, whereas the notion of ‘by partitioning’ can be implemented by either the technique of bisection (ART‐B) or the technique of random partitioning (ART‐RP). In this paper, we propose a generic approach that can be used to implement different notions. In the new approach, test cases are simply selected based on test profiles that are in turn designed according to certain notions. In this study, we design several test profiles for the notions of ‘by exclusion’ and ‘by partitioning’, and then use these profiles to illustrate our new approach. Our experimental results show that compared with the original RRT, ART‐B, and ART‐RP algorithms, our new approach normally brings at least a higher failure‐detection capability or a lower computational overhead. Copyright © 2011 John Wiley & Sons, Ltd.

Given a set of spatial points $D$ containing keywords information, a set of query objects Q and m query keywords, a top-k aggregate nearest keyword (ANK) query retrieves k objects from Q with the minimum sum of distances to its nearest points in D such that each nearest point matches at least one of query keywords. For example, consider there is a spatial database D which manages facilities (e.g., school, restaurants, hospital, etc.) represented by sets of keywords. A user may want to rank a set of locations with respect to the sum of distances to nearest interested facilities. For processing this query, several algorithms are proposed using IR2-Tree as index structure. Experiments on real data sets indicate that our approach is scalable and efficient in reducing query response time.

Peer-to-peer (P2P) traffic is arguably the biggest contributor to today's Internet traffic. So far, very limited effort has been paid to theoretically analyze the effect of bringing locality-awareness into P2P applications. In this paper, we analyze three locality-awareness policies for BitTorrent-like system: tracker locality(neighbor selection), choker locality(choking and unchoking) and picker locality(piece picking). Based on an H-sphere model which captures the essential property of Internet topology, we analyze how much network load saving can be expected for these locality policies, as well as their impact to the downloading efficiency of the system. Through analytical study and simulation, we have the following findings. First, when the system enters the steady state, the locality policies proposed in this paper can achieve the same downloading efficiency as the standard BitTorrent system. Second, these locality policies can significantly alleviate the traffic load on the Internet. Finally, we find that there exists no all-around winner. Instead, each policy performs the best under different system configuration parameters, such as number of unchoked peers.

Given a data point set D, a query point set Q and an integer k, the group nearest group (GNG) query finds a subset of points from D, omega (|omega| les k), such that the total distance from all points in Q to the nearest point in omega is no greater than any other subset of points in D, omega(|omega| les k). GNG query can be found in many real applications. In this paper, exhaustive hierarchical combination algorithm (EHC) and subset hierarchical refinement algorithm (SHR) are developed for GNG query processing. The superiority of SHR in terms of efficiency and quality compared to existing algorithms developed originally for data clustering is demonstrated.

Traditional testing methods tend to generate test suite for programs of which parameters are basic data types, e.g. integer, real or Boolean. However, there are lots of programs with inputs of complex data types, e.g. array, point or structure, traditional testing methods canpsilat deal with them. This paper focuses on testing programs with array parameter that is widely used in programming. The difficulties to generate array test suite lie in two reasons. The first reason is that array is composed by several ordered, interactional elements. Thus, we cannot take the array as several independent, isolated elements when generating array test suite. The second one is that we should consider the implicative attributes of array effecting the generation of test suite, e.g. the size of array, or some special elements in array. This paper proposes a new testing method, extended combination testing (ECT), to generate array test suite systemically. Through based on traditional combination testing, ECT extends its application scope form basic data test suite generation to array test suite generation. ECT proceeds in three steps: (1) yield values of other non-array parameters and spawn them to a proper combination; (2) identify the implicative attributes of array; (3) choose values for each array elements, and combine values of array elements with proper combination coverage into test suite. While in traditional combination test, only step (3) is included. The results of our experiment validate the effectiveness of ECT in generating array test suite.

Mobile sensors are widely deployed in Wireless Sensor Networks (WSNs) to satisfy emerging application requirements. Specifically, processing location dependent queries in mobile WSNs is still a challenging problem due to sensor mobility. We present an Event-based Location Dependent Query (ELDQ) model that continuously aggregate data in specific areas around mobile sensors of interests to provide event-based location dependent data services to the users. ELDQs generalize several typical query types and are important in many applications. However, existing approaches are incapable of efficiently answering ELDQs. In this paper, we propose a set of techniques to process ELDQs while optimizing system performance. Cost analysis and simulation results indicate that our techniques greatly reduce the cost of processing ELDQs while achieving relatively high accuracy and short response time.

This paper presents the measurement study of locality-aware P2P solutions over real-world Internet autonomous systems (AS) topology. By using the accesses of nodes of PlanetLab testbed, we create a detailed AS-level map including the end-to-end path of all nodes, as well as the relationship of all involved ASes. Based on this map, we evaluate the performance of a set of locality-aware P2P solutions, including an optimal solution guaranteeing the minimum AS hop count, as well as modified BitTorrent system with locality-awareness built into its neighbor selection, peer choking/unchoking, and piece selection processes. Our findings suggest that locality-awareness can help existing P2P solution to significantly decrease load on Internet, and achieve shorter downloading time. By comparing the performance of different kinds of locality-aware and traditional BitTorrent systems, we also point out the necessity to tradeoff between the goals of optimizing AS-related performance and achieving fairness among peers such as intra-AS traffic and peer burden fairness.

Component-based software engineering (CBSE) has been proven to be effective in the development and management of distributed software systems. Currently, some challenges appear when building time-critical applications for the distributed systems based on current component technologies. The remaining issues that components themselves have not undertaken are the two key characteristics of distributed real-time systems: timing properties and dynamic properties. To solve the issues and facilitate the development of distributed real-time applications in client-server computing pattern, this paper proposes a kind of reflective design for component-based distributed systems with timing constraints. At the end, an example is given to validate the design. Experimental results show that an application’s timing properties can be met by adopting this design approach and dynamic properties have been achieved by dynamically adjusting components on demand.

Retrieving related graphs containing a query graph from a large graph database is a key issue in many graph-based applications, such as drug discovery and structural pattern recognition. Because sub-graph isomorphism is a NP-complete problem [4], we have to employ a filter-and-verification framework to speed up the search efficiency, that is, using an effective and efficient pruning strategy to filter out the false positives (graphs that are not possible in the results) as many as possible first, then validating the remaining candidates by subgraph isomorphism checking. In this paper, we propose a novel filtering method, a spectral encoding method, i.e. GCoding. Specifically, we assign a signature to each vertex based on its local structures. Then, we generate a spectral graph code by combining all vertex signatures in a graph. Based on spectral graph codes, we derive a necessary condition for sub-graph isomorphism. Then we propose two pruning rules for sub-graph search problem, and prove that they satisfy the no-false-negative requirement (no dismissal in answers). Since graph codes are in numerical space, we take this advantage and conduct efficient filtering over graph codes. Extensive experiments show that GCoding outperforms existing counterpart methods. 1.

This paper describes a method to automatically acquire query translation pairs by mining web click-through data. The extraction requires no crawling or Chinese words segmentation, and can capture popular translations. Experimental results on a real click-through data show that only 17.4% of the extracted queries are in the dictionary, and our method can achieve 62.2% (in top-1) to 80.0% (in top-5) precision in translating web queries. Moreover, the extracted translations are semantically relevant to the source query, which is particularly useful for Cross-Lingual Information Retrieval (CLIR).

Software testing is a practical activity combined with theory, technology, tool, and management. Assistant tool for testing plays an important role in software development in practice, and should not be neglected. Taking the popularly used C and C+ + programs as the test objects, this paper explores how to implement a semi-automatic tool (named CppTest) with the testing capability in three levels, namely (1) structural testing in method level, (2) state-based class level testing through modeling state transition behaviors using an extended finite state machine (EFSM), and (3) system level black-box testing with some traditional strategies. During the later debugging stage, the prototype system can perform clustering analysis on the failure executions and sample fairly few representative test executions to direct program diagnosis. CppTest mainly possesses the merits as follows: (1) multi-levels' testing support, (2) high automation and visualization, (3) test project management, and (A) favorable scalability. In addition, some future research directions are also explored

Component-based software (CBS) is built through the ways of composition and integration, this development style is quite favorable for modification or upgrade of system. However, making sure that the modified system still works is a challenge. In order to facilitate the maintenance activities and their optimum configuration, it is necessary to measure the changes. This paper mainly discusses the case of component modification, including the single component change and the changes of multi-components, and presents the algorithms for measuring the change impacts in CBS for each situation. Finally, the on-going research plan is addressed.

Constrained multi-objective optimization problems (CMOP) are challenging and difficult to solve. In this paper, a simple and practical evolutionary algorithm for constrained multi-objective optimization problems (EACMOP) is presented, by defining constraints using non-parameter punitive functions, using Pareto strength value to represent Pareto order strength among individuals and using crowding density to ensure group diversity. It defines the evolutionary algorithm fitness functions by combining constraint treatment, comparison of Pareto strength optimization and crowding density. Test results on several benchmark functions showed that the approach is effective and robust

The integration of grid database is a hot researching field of grid database management nowadays. The middleware of data access and integration then plays an important role in grid database integration. This paper discusses the integration and the access of grid database based on the research and the analysis of OSGA-DAI framework and its internal procedure. We put forward a system framework and model design to extend and improve OGSA-DAI functions. The system presents good performances for the access and the integration of grid database. Finally, combined with the relative key technology and programming skills, we give out a solution of the access and the integration of grid database

As XML has become the standard format for representing structured and semi-structured data on the Web, the methods for designing XML schemas is becoming more and more important. XML schemas represent the logical models of the documents. In order to design or integrate XML schemas, it is necessary to first design the conceptual structures with a proper conceptual model. We thus specify a XML conceptual model, XUML, which has following characteristics comparing with the existing XML conceptual models: 1) expressing the containment semantics more explicitly; 2) supporting the concept of Business Components; 3) specifying the data dependencies in multiple contexts. XUML is defined based on the UML2 standard, so it will be more friendly and practical for those who have had knowledge and experience on UML. Based on XUML, we further present a framework of the methodology which is dedicated to the design of XML documents and XML databases. In this paper, the focus is put on XUML.

In recent years, component-based software has been widely used in various application domains and becomes a fairly popular software form. However, due to the lack of information about the externally-developed components, system testers (i.e., component users) generally can't perform effective testing (especially regression testing) on their component-based systems. Component users don't know the details about change in component, so they aren't able to select the proper test cases to retest the modified system. In this paper, we present an improved regression testing method based on the enhanced change information of component version to test the software system containing some modified components. It is a collaborative testing method, needing the joint participations of component developer and user. Component developers calculate the change information from labeled method call graph and provide it to component users via XML files. Component users use this change information and their instrumentation records together to pick out test cases for next-round testing. In addition, we have employed preliminary experiments on some medium scale systems, the experiment results show that our regression testing method is fairly feasible and cost-effective in practice.

The government information portal is a very important part of the e-government. With the development of the society and the e-government, modern governments put forward higher level requirements on the e-government. Grid technology provides strong technology support for it. Grid technology researching has great practical value in the government information portal application. This paper analyzed the deficiency in the existing government information portal and discussed the flows of information collecting, storing, processing and displaying in the information grid environment based on grid technology. Building an electronic government information portal framework using OGSA environment, we have implemented a prototype system, which is used for government information integration and information displaying. The system takes grid service as the basic unit and applies the function of grid environment to fulfill the systems integration and government information sharing.

Hard real-time systems require predictable performance despite the occurrence of failures. In this paper, the authors proposed a new fault-tolerant priority assignment algorithm based on worst-case response time schedulability analysis for fault-tolerant hard real-time system. This algorithm can be used, together with the schedulability analysis, to effectively improve system fault resilience when the two traditional fault-tolerant priority assignment policies cannot improve system fault resilience. Also, a fault-tolerant priority configuration search algorithm for the proposed analysis was presented.

Exception handling is a powerful mechanism that separates the error handling code from normal code. It makes software do its utmost to run on the normal state. However, incorrect usage of exception will bring about more potential faults in the code. Based on the study of exception model of C++, a typical object-oriented language, we present a method of constructing control flows for both explicit exception and implicit exception. Subsequently, provide a model that can improve the robustness of programs via static exception analysis. This paper also discusses methods of dynamic exception structural testing and proposes several testing strategies. Finally, a rough prototype tool for exception analysis and testing has been implemented, and some preliminary experiments have been performed under it. The experiment results show that our methods of static exception inspection and dynamic exception testing are fairly effective to expose the potential exception errors. Furthermore, statistical structural exception testing is also considerably useful to find out some normal faults.

The main attack strategy in an intense robotic soccer match is for an attack robot (AR) to avoid the competing side's threat and to avoid the most threatening defensive robot of the opponent to reach the objective and to effectively initiate an attack with the help of the cooperation robot (CR). This paper defines an attacking model and a cooperative model and their algorithms. Two co-evolution computation (CEC) populations of robot are also designed: one is denoted as AR subset, the other CR subset. Based on this definition, the paper proposes a new multiple robots avoidance based on CEC method (or MRACEC for short). A theoretical analysis indicates that the MRACEC method has better robustness and optimizing ability.

Accurately and effectively detecting the locations where search queries are truly about has huge potential impact on increasing search relevance. In this paper, we define a search query's dominant location (QDL) and propose a solution to correctly detect it. QDL is geographical location(s) associated with a query in collective human knowledge, i.e., one or few prominent locations agreed by majority of people who know the answer to the query. QDL is a subjective and collective attribute of search queries and we are able to detect QDLs from both queries containing geographical location names and queries not containing them. The key challenges to QDL detection include false positive suppression (not all contained location names in queries mean geographical locations), and detecting implied locations by the context of the query. In our solution, a query is recursively broken into atomic tokens according to its most popular web usage for reducing false positives. If we do not find a dominant location in this step, we mine the top search results and/or query logs (with different approaches discussed in this paper) to discover implicit query locations. Our large-scale experiments on recent MSN Search queries show that our query location detection solution has consistent high accuracy for all query frequency ranges.