Vol. 37 · No. 3

Acceleration of stencil computation can be effectively improved by utilizing the memory resource. In this paper, in order to reduce the branch divergence of traditional mapping method between the global memory and the shared memory, we devise a new mapping mechanism in which the conditional statements loading the boundary stencil computation points in every XY-tile are removed by aligning ghost zone to reduce the synchronization overhead. In addition, we make full use of single XY-tile loaded into registers in every stencil computation point, common sub-expression elimination and software prefetching to reduce overhead. At last detailed performance evaluation demonstrates our optimized policies are close to optimal in terms of memory bandwidth utilization and achieve higher performance of stencil computation.

The concept of service in Service-Oriented Architecture (SOA) makes possible to introduce other ideas like service composition, governance and virtualization. Each of these ideas, when exercised to an enterprise level, provides benefits in terms of cost and performance. These ideas bring many new opportunities for the project managers in making the estimates of effort required to produce SOA systems. This is because the SOA systems are different from traditional software projects and there is a lack of efficient metrics and models for providing a high level of confidence in effort estimation. Thus, in this paper, an efficient estimation methodology has been presented based on analyzing the development phases of past SOA based software systems. The objective of this paper is twofold: first, to study and analyze the development phases of some past SOA based systems; second, to propose estimation metrics based on these analyzed parameters. The proposed methodology is facilitated from the use of four regression(s) based estimation models. The validation of the proposed methodology is cross checked by comparing the predictive accuracy, using some commonly used performance measurement indicators and box-plots evaluation. The evaluation results of the study (using industrial data collected from 10 SOA based software systems) show that the effort estimates obtained using the multiple linear regression model are more accurate and indicate an improvement in performance than the other used regression models.

Concept similarity measure aims at identifying a degree of commonality of two given concepts and is often regarded as a generalization of the classical reasoning problem of equivalence. That is, any two concepts are equivalent if and only if their similarity degree is one. However, existing measures are often devised based on objective factors, e.g. structural-based measures and interpretation-based measures. When these measures are employed to characterize similar concepts in an ontology, they may lead to unintuitive results. In this work, we introduce a new notion called concept similarity measure under preference profile with a set of formally defined properties in Description Logics. This new notion may be interpreted as measuring the similarity of two concepts under subjective factors (e.g. the agent's preferences and domain-dependent knowledge). We also develop a measure of the proposed notion and show that our measure satisfies all desirable properties. Two algorithmic procedures are introduced for top-down and bottom-up implementation, respectively, and their computational complexities are intensively studied. Finally, the paper discusses the usefulness of the approach to potential use cases.

In this paper, we consider the problem of parameter selection and uncertainty measurement for a variable precision probabilistic rough set. Firstly, within the framework of the variable precision probabilistic rough set model, the relative discernibility of a variable precision rough set in probabilistic approximation space is discussed, and the conditions that make precision parameters α discernible in a variable precision probabilistic rough set are put forward. Concurrently, we consider the lack of predictability of precision parameters in a variable precision probabilistic rough set, and we propose a systematic threshold selection method based on relative discernibility of sets, using the concept of relative discernibility in probabilistic approximation space. Furthermore, a numerical example is applied to test the validity of the proposed method in this paper. Secondly, we discuss the problem of uncertainty measurement for the variable precision probabilistic rough set. The concept of classical fuzzy entropy is introduced into probabilistic approximation space, and the uncertain information that comes from approximation space and the approximated objects is fully considered. Then, an axiomatic approach is established for uncertainty measurement in a variable precision probabilistic rough set, and several related interesting properties are also discussed. Thirdly, we study the attribute reduction for the variable precision probabilistic rough set. The definition of reduction and its characteristic theorems are given for the variable precision probabilistic rough set. The main contribution of this paper is twofold. One is to propose a method of parameter selection for a variable precision probabilistic rough set. Another is to present a new approach to measurement uncertainty and the method of attribute reduction for a variable precision probabilistic rough set.

This paper is in the framework of the specification and verification of concurrent dynamic systems. For this purpose we propose the model of Refinable Recursive Petri Nets (RRPN) under a maximality semantics. In this model a notion of undefined transitions is considered. The underlying semantics model is the Maximality Abstract Labeled Transition System (AMLTS). Then, the model supports a definition of a hierarchical design methodology. The example of a cutting flame machine is used for illustrating the approach.

The travelling salesman problem (TSP), a famous NP-hard combinatorial optimisation problem (COP), consists of finding a minimum length tour that visits n cities exactly once and comes back to the starting city. This paper presents a resolution of the TSP using the breakout local search metaheuristic algorithm (BLS), which is based on the iterated local search (ILS) framework and improves it by introducing some fundamental features of several well-established metaheuristics such as tabu search (TS) and variable neighbourhood search (VNS). BLS moves from a local optimum of a neighbourhood to another by applying perturbation jumps whose type and number are determined adaptively. It has already been applied to many COP and gives good results. This innovative hybridisation resolved well 41 instances from the commonly used benchmark library TSPLIB. The high quality of experimental results shows the competitiveness of the proposed algorithm compared to other algorithms based on local search.

Usually people prefer to focus on creative rather than repetitive and schematic work patterns. Still, they must spend a lot of time complying with the procedures, selecting the information they receive and repeatedly restoring the previous state of work. This paper proposes the Mobile INteractive Document architecture (MIND) – a document-centric uniform interface to provide both effective communication of content and coordination of activities performed on documents. MIND documents are proactive, capable of initiating process activities, interacting with individuals on their personal devices and migrating on their own between collaborators. Each MIND document is a mobile agent that has built-in migration policy to control its own workflow and services enabling proper processing of contained information. The architecture supports users in the implementation of procedures, and selection of services needed to work on the document. A Personal Document-Agent (PDA) is a further development of MIND aimed at preserving continuity of state of individuals' work to support their creativity and comfort of their daily work.

With the coming up of plethora of web applications and technologies like sensors, IoT, cloud computing, etc., the data generation resources have increased exponentially. Stream processing requires real time analytics of data in motion and that too in a single pass. This paper proposes a framework for hourly analysis of streamed data using Bloom filter, a probabilistic data structure where hashing is done by using a combination of double hashing and partition hashing; leading to less inter-hash function collision and decreased computational overhead. When size of incoming data is not known, use of Static Bloom filter leads to high collision rate if data flow is too much, and wastage of storage space if data is less. In such cases it is difficult to determine the optimal Bloom filter parameters (m, k) in advance, thus a target threshold for false positives (f p) cannot be guaranteed. To accommodate the growing data size, one of the major requirements in Bloom filter is that filter size m should grow dynamically. For predicting the array size of Bloom filter Kalman filter has been used. It has been experimentally proved that proposed Adaptable Bloom Filter (ATBF) efficiently performs peak hour analysis, server utilization and reduces the time and space required for querying dynamic datasets.

Visual tracking of a deformable object is a challenging problem, as the target object frequently changes its attributes like shape, posture, color and so on. In this work, we propose a model-free tracker using clustering to track a target object which poses deformations and rotations. Clustering is applied to segment the tracked object into several independent components and the discriminative parts are tracked to locate the object. The proposed technique segments the target object into independent components using data clustering techniques and then tracks by finding corresponding clusters. Particle filters method is incorporated to improve the accuracy of the proposed technique. Experiments are carried out with several standard data sets, and results demonstrate comparable performance to the state-of-the-art visual tracking methods.

This paper deals with mining conversational content from the social media. It focused on two issues: opinion and emotion classification and identification of authoritative reviewers. The paper also describes applications representing the results obtained in the given areas. Authority identification can be used by organizations to search for experts in their specific areas to employ them. The opinion and emotion analysis can be useful for providing decision-making support.

Classification of automatically generated malware is an active research area. The amount of new malware is growing exponentially and since manual investigation is not possible, automated malware classification is necessary. In this paper, we present a static malware detection system for the detection of unknown malicious programs which is based on combination of the weighted k-nearest neighbors classifier and the statistical scoring technique from [merkel2010statistical]. We have extracted the most relevant features from portable executable (PE) file format using gain ratio and have designed a heterogeneous distance function that can handle both linear and nominal features. Our proposed detection method was evaluated on a dataset with tens of thousands of malicious and benign samples and the experimental results show that the accuracy of our classifier is 98.80