Abstract:SIMD extension is an acceleration component integrated into the general processor for developing data level parallelism in multimedia and scientific computing applications. Firstly, in this study the background and research status of SIMD extension are introduced. Next, challenges and latest research achievements in SIMD auto-vectorization are discussed from three perspectives: development method, data layout and vectorization in multi-platform. Finally, some future trends in the SIMD compiling optimization are addressed.

Abstract:Software configuration has been a big challenge due to the diversity, complexity, flexibility and customizability of a system. Configuration error has become a dominant cause leading to system failure and service outage. To improve software availability and reliability, many researchers and institutions have devoted their efforts on software misconfiguration troubleshooting. This paper first builds an analytical framework with establishment of 3 aspects and multiple perspectives, covering the method type, style and applicability. Based on this framework, the paper provides an overview of the state of art of misconfiguration troubleshooting along with analysis on the current leading methods of software misconfiguration troubleshooting. Finally, this paper summarizes the shortcomings in the current research and outlinesthe development prospects of future research. This paper aims to provide some available information and beneficial insight for future researches.

Abstract:As one of the most important essentials of CPU, mathematical function libraries play a key role in scientific and engineering computing with high performance computers. Existing testing techniques and platforms can only evaluate function libraries from one or two aspects, therefore are unable to provide an evaluation result as a whole picture. Consequently, they are applicable for a specific targeting architecture and the scalability is restricted. To address this problem, this study proposes a novel testing platform BMltest (Basic math library test). It constructs the testing suite, which is composed of eigenvalues, IEEE-754 special values and IEEE-754 normalized values, to improve the cover rate of the floating numbers. A MPFR (multiple-precision floating-point reliable library) based precision test is introduced, and as a result, the reliability is improved. A code isolation based performance test is also described, so as to further eliminate the impact from enclosing circumstance. Some practical evaluating strategies are proposed to evaluate the test result. Such design makes the testing suite not correlated to mathematical functions, thereby ensuring the applicability. The experimental results show that, by testing 855 functions from various libraries, including GNU, Open64 and Mlib, the testing suite provided by BMltest is more efficient and the precision test is more reliable. At the same time, compared with those of other testing platforms, the performance test is more stable.

Abstract:Anti-Reversing protection for persistent and high-insensitive software core algorithm has become an insistent demand for the research of software security and even for the whole software industry. Virtual machine based software protection has been widely used to protect the core algorithm from being reversed, but it is not sufficient for the current method to defend against cumulative attack and thus cannot provide long-term effective protection. Time diversity is used to fight against cumulative attack to allow software to execute along variant paths in different running time. A virtual machine based software protection method with time diversity, called TDVMP, is proposed in the paper. The key idea of the method is to construct multiple execution paths with equivalent semantics leading to dynamically variant execution paths in running time. Main design issues of TDVMP, such as construction and selection of multiple execution paths, are discussed in detail. Furthermore, a metric named variation of execution paths to evaluate the effectiveness of time diversity is proposed, and the methods to measure and compute the metric are also presented. A prototype of TDVMP is implemented, and upon which the experiments are carried out with a set of practical use cases. Experiment results show that TDVMP is effective and applicable for core algorithm anti-reversing protection.

Abstract:Generic programming has emerged as a paradigm for the development of highly reusable and safe software libraries. Generic constraints mechanism includes a collection of features for constraining generic parameters and verification of the validity of generic parameter instantiated, thereby guarantees dependability and safety of generic programs. This paper first reviews the current research status of generic constraints, exposing the difficulty of describing and verifying generic programs with dynamic semantic constraints. Based on a new description of generic constraints of Apla language, it then proposes three main types of generic constraints mechanism: constraints of basic data types, constraints of custom abstract data types and constraints of subroutines. Next, with the help of Isabelle theorem prover, the paper designs the generic constraints matching detection and validation algorithms and further gives the implementation scheme of generic constraints mechanism in PAR platform. It confirms that the proposed generic constraints mechanism can solve a series of complex generic constraints problems, and so markedly improves dependability and safety of generic programs.

Abstract:Social recommender systems have recently become one of the hottest topics in the domain of recommender systems. The main task of social recommender system is to alleviate data sparsity and cold-start problems, and improve its performance utilizing users' social attributes. This paper presents an overview of the field of social recommender systems, including trust inference algorithms, key techniques and typical applications. The prospects for future development and suggestions for possible extensions are also discussed.

Abstract:In modeling a discrete event system, the map from physical system to logic system may be not complete due to the complex behaviors of the system. In this paper, the concept of incomplete model is introduced. Next, a method for judging diagnosability in incomplete model is proposed, and the corresponding on-line version of the method is also presented. Incomplete behaviors can be added to the model, and makes the model complete. Offline diagnosability is judged by classical twin-plant method. According to the ordered observations and prefix of language, the incomplete behaviors are judged and disposed. With an additional method that judge incomplete behaviors, the incomplete behaviors are added into model, and the online diagnosability is judged incrementally in twin-plant by incomplete behaviors. By judging diagnosability online, whether a fault can be found exclusively by limit observations is decided. The proposed methods suit for the systems which is discrete.

Abstract:Low-resolution is an important issue when handling real world image recognition problems. The performance of traditional recognition algorithms, e.g. LDA/PCA, usually drops drastically due to the loss of discriminant information compared to those for high-resolution or super-resolution images. In order to solve this problem, many methods have been proposed in recent years based on coupled projections, i.e. learning two sets of different projections, one for high-resolution images and one for low-resolution images. For example, SDA (simultaneous discriminant analysis) obtains projections by maximizing the average between-class scatter while minimizing the average within-class scatters. Like LDA, SDA cannot separate projected classes, especially for those that are closer to each other. In this paper, a novel discriminant analysis method is proposed to achieve the optimal projections by maximizing the minimum distance between pair-wise classes. Experiments on several image datasets verify the efficiency of the presented methods.

Abstract:As one of the effective methods to ease the information overload problem, recommender systems have become extremely popular in social media. However, recommender methods suffer from the cold-start problems in new item recommendations and new user recommendations. To combat the cold-start problems in new item recommendations, the concept of user time weights is proposed to characterize the time interval between the user evaluating time and item distributing time. According to the weights, it can determine whether the user is a positive user or a negative user, and the degree of the user's preference for new items. Tripartite graphs are used to picture relations among user-item-tag, and user-item-attribute. Combing information among users, tags, attributes of items and time weights, functions for predicting the rating are defined and a new personalized recommendation algorithm is constructed. Overall experimental results show that the proposed method not only brings satisfied personalized items but also pleasantly surprises different users with different preferences. Comparative experiments illustrate the proposed method is much higher in accuracy and novelty. Cross-validation experiments demonstrate that the new method is effective to solve the cold-start problem in new item recommendations.

Abstract:Corpus-based speech synthesis is the most widely-used speech synthesis technology in home and abroad. In this type of synthesis method, unit selection is crucial to the speech synthesis. This paper designs a system model of Tibetan speech synthesis by analyzing the attributive characteristics of Tibetan text, and presents a mixed units mode with Tibetan components, combinational components, characters, words and sentences. The method effectively preserves the integrity of larger units and the flexibility and robustness of small units. At the same time, it provides the unit selection strategies and algorithms of Tibetan speech synthesis. Experimental data indicates that the strategies and algorithms are effective and the coverage of units reaches expected target in both the open corpus and closed corpus.

Abstract:Taking entity as the basic unit to organize the tuples in query processing is an effective way of managing low-quality data. As many descriptions of attribute value in an entity, join operator must support similarity join over multiple values. Entity similarity join is more effective than traditional similarity join in data cleaning, information integration, fuzzy keyword search, fraud detection, and text aggregation. In this paper, an entity similarity join algorithm, ES-JOIN, is designed by adopting the structure of the double layer prefix index. The presented method is suitable for solving set similarity join problem based on fuzzy elements matching, and thus is a better choice than the traditional set similarity join which only considers exact element match. In order to accelerate the join process, a new filtering measures is proposed to optimize the algorithm, and an optimization algorithm, OPT_ES-join, is also obtained. Experiments demonstrate that the ES-JOIN algorithm has good efficiency and scalability, and the filter measures is very effective.

Abstract:The exponential growth of data has posed serious challenges to the data management and analysis. Join query is a common data analysis operation, and MapReduce is a programming model implemented for parallel processing on large-scale datasets. Therefore the research on MapReduce based join algorithms and its cost model has a certain academic significance and application value. This study believes that the I/O (including the network and the local I/O) cost is the main factor affecting the performance of MapReduce based join algorithm. Furthermore, as the I/O cost is determined by the feature of both datasets and join operation, the executed plan of multi-ways join could be optimized by evaluating the I/O cost of two-ways join. In the study, an I/O cost model of two-ways join is proposed and then formally defined as a simple extension to the existing MapReduce based join algorithms, resulting in six join algorithms and their I/O cost functions through write-box analysis. In addition, an selection algorithm to find the best executed plan of multi-ways join is presented. The correctness and accuracy of the I/O cost model are validated through a series of experiments. The experiment results suggest that the I/O cost can accurately reflect the algorithm performance.

Abstract:In the research of privacy preserving transaction data publishing, the existing methods are always designed for the centralized structure. The paper proposes a differential privacy publishing strategy to protect data privacy and maximize utility of the output data in the distributed environment. The new method combines the utility optimization of the output with differential privacy constraints and builds a distributed nonlinear programming model. Furthermore, two solutions based on global and local data respectively are designed to solve the distributed model securely. As shown in the theoretical analysis and the experimental results, the publishing strategy can achieve significant improvements in terms of privacy, security, and applicability.

Abstract:In microblog networks, the propagation of messages is closely related to the influence of opinion leaders. However, it is hard to quantitate the influence of opinion leaders in different types of messages, which bring great challenges to evaluate the influence of opinion leaders and predict the propagation trend of a message. To solve this problem, this paper proposes a method to measure and analyze opinion leaders' influence in microblog based on the dynamical process of message propagation. By studying the spread patterns of messages, dynamical directed graph is employed to model the propagation process. The propagation of a message can be decomposed into sub-propagations raised by opinion leaders, which can be described as exponential truncated power-law decay function. By estimating the parameters in the model, the initial influence, influence decay rate and influence insistency of opinion leaders can be evaluated quantitatively. Results of the experiment based on the data collected from Sina microblog, show the total retweeted messages is weakly correlated to the number of opinion leaders in the spread process. In addition, the number of followers possessed by opinion leaders is positively correlated to the power of their initial influence. However, it exhibits no correlation with the decay rate and insistence time of the influence. The efficiency and correctness of the proposed model are validated by predicting the spread trends of hot messages in actual microblog networks, which is important to network marketing and message propagation controls.

Abstract:Bridge operation safety is critical to national security and people's livelihood. The structural health monitoring of bridges has emerged as an increasingly active research area. Wireless sensor networks (WSNs) technology is known to be easy to deploy and inexpensive to maintain. It is thus suitable for structural health monitoring of bridges. This paper gives a survey on structural health monitoring systems based on WSNs technology. Some basic theories and typical methods in subsystems are presented, and critical technologies are analyzed. At the end, varies issues in existing systems and directions on future work are analyzed and summarized.

Abstract:Vehicular ad hoc networks (VANETs) are characterized by high mobility of vehicle nodes, intermittent connectivity, and rapid dynamic topology. Data to be disseminated can hardly be kept within a target area, and therefore cannot provide continuous services for the vehicles passing by. Motivated by the fact that there are large amounts of roadside parked vehicles in urban areas, this paper proposes a parked-vehicle assisted data dissemination (PADD) paradigm for vehicular ad hoc networks (VANETs). In PADD, the parked vehicles in the target area are managed on the basis of parking cluster, the data to be disseminated are forwarded from the data source to the selected parking clusters, and the pub/sub scheme is adopted to achieve data dissemination within one hop of the parking cluster. Theoretical results illustrate the effectiveness of the proposed approach, and meanwhile simulation results based on a real city map and realistic traffic situations show that PADD achieves a higher delivery ratio with lower network load and reasonable transmission delay.

Abstract:Free-riding destroy the foundation of BitTorrent file sharing, and result in bad system performance. The choking scheme adopted in BitTorrent nowadays can suppress the free-riding, however the coexisting unchoking scheme in which the random peers are chosen lend opportunities to free-riders. An unchoking scheme based on probabilistic link exchange, or PLX for short, is provided in this paper. The new scheme can suppress the free-riding effectively while guaranteeing the unchoking function. Free-riders can't enter into the system because they will not be unchoked by PLX after the links to free-riders are choked as PLX works based on link exchanges. Furthermore, by virtue of the mathematical designs for the probability of the link exchange, PLX can distinguish the contribution of peers, adjust their location in network according to contribution, and improve the fairness of the system. The in-depth theoretical analyses and experimental evaluations show that comparing with other methods for fighting against free-riding attacks, the PLX unchoking scheme is simple, direct and effective..

Abstract:Policy refinement is an important technology to resolve the configuration complexity of access control policies in distributed applications. Existing methods for policy refinement describe and refine policies layer by layer. However, they are weak in dealing with the relationship between policies. In this study, policies and the relationship between them are described based on the policy refinement tree where policies conflict analysis is performed on the leaf nodes to allow using R-refutation calculus of open logic to analyze refinement policy correlation properties. This method can resolve conflicting policies while correctly maintaining mutual exclusion, combination, access path coordination, and refinement mapping of policies. It can also resolve conflicting policies of different types in order, and freely make a choice among conflicting policies. Experiments and performance analysis demonstrate that the presented method meets the need of dynamic adaption of policy refinement for service-oriented application systems on SaaS platform.