Vol. 31 · No. 6

A major aim of the Web service platform is the integration of existing software and information systems. Data integration is a central aspect in this context. Traditional techniques for information and data transformation are, however, not sufficient to provide flexible and automatable data integration solutions for Web and Cloud service-enabled information systems. The difficulties arise from a high degree of complexity in data structures in many applications and from the additional problem of heterogeneity of data representation in applications that often cross organisational boundaries. We present an integration technique that embeds a declarative data transformation technique based on semantic data models as a mediator service into a Web service-oriented information system architecture. Automation through consistency-oriented semantic data models and flexibility through modular declarative data transformations are the key enablers of the approach. Automation is needed to enable dynamic integration and composition. Modifiability is another aim here that benefits from consistency and modularity.

The distributed control system of the Large Hadron Collider (LHC) presents many challenges due to its inherent heterogeneity and highly dynamic nature. One critical challenge is providing access control guarantees within the middleware. Role-based access control (RBAC) is a good candidate to provide access control. However, in an equipment control system transactions are often dependent on user context and device context. Unfortunately, classic RBAC cannot be used to handle the above requirements. In this paper we present an extended role-based access control model called CMW-RBAC. This new model incorporates the advantages of role-based permission administration together with a fine-grained control of dynamic context attributes. We also propose a new technique called dynamic authorization that allows phased introduction of access control in large distributed systems. This paper also describes motivation of the project, requirements, and overview of its main components: authentication and authorization.

This paper sumarizes the data preprocessing methods used for modelling of laser beam impact. At the beginning, it gives a short overview of physical background of interaction between a laser beam and processed material. Also the input data sets are described and the way of their acquisition is shown, so that the reader could get a good idea of the whole problem. The main content of the paper is formed by description of the data preparation process that consists of several steps – automatic heat-affected area detection, samples parameterization and their surface approximation. In all cases our methods with the best results are described in a detail. At the end of the paper, our results are summarized, discussed and possible plans for the future research are introduced.

With emergence of new techniques, data in many fields are getting larger and larger, especially in dimensionality aspect. The high dimensionality of data may pose great challenges to traditional learning algorithms. In fact, many of features in large volume of data are redundant and noisy. Their presence not only degrades the performance of learning algorithms, but also confuses end-users in the post-analysis process. Thus, it is necessary to eliminate irrelevant features from data before being fed into learning algorithms. Currently, many endeavors have been attempted in this field and many outstanding feature selection methods have been developed. Among different evaluation criteria, mutual information has also been widely used in feature selection because of its good capability of quantifying uncertainty of features in classification tasks. However, the mutual information estimated on the whole dataset can not exactly represent the correlation between features. To cope with this issue, in this paper we firstly re-estimate mutual information on identified instances dynamically, and then introduce a new feature selection method based on conditional mutual information. Performance evaluations on sixteen UCI datasets show that our proposed method achieves comparable performance to other well-established feature selection algorithms in most cases.

A longest increasing subsequence problem (LIS) is a well-known combinatorial problem with applications mainly in bioinformatics, where it is used in various projects on DNA sequences. Recently, a number of generalisations of this problem were proposed. One of them is to find an LIS among all fixed-size windows of the input sequence (LISW). We propose an algorithm for the LISW problem based on cover representation of the sequence that outperforms the existing methods for some class of the input sequences.

Pareto efficiency is a seminal condition in the bargaining problem which leads autonomous agents to a Nash-equilibrium. This paper investigates the problem of the generating Pareto-optimal offers in bilateral multi-issues negotiation where an agent has incomplete information and the other one has perfect information. To this end, at first, the bilateral negotiation is modeled by split the pie game and alternating-offer protocol. Then, the properties of the Pareto-optimal offers are investigated. Finally, based on properties of the Pareto-optimal offers, an algorithmic solution for generating near-optimal offers with incomplete information is presented. The agent with incomplete information generates near-optimal offers in O(n łog n). The results indicate that, in the early rounds of the negotiation, the agent with incomplete information can generate near-optimal offers, but as time passes the agent can learn its opponents preferences and generate Pareto-optimal offers. The empirical analysis also indicates that the proposed algorithm outperform the smart random trade-offs (SRT) algorithm.

This paper presents a new load-balancing algorithm for shared memory multiprocessors that is currently being applied to the parallel simulation of logic circuits, specifically VHDL simulations. The main idea of this load-balancing algorithm is based on the exploitation of the usual characteristics of these simulations, that is, cyclicity and predictability, to obtain a good load balance while preserving the locality of references. This algorithm is useful not only in the area of logic circuit simulation but also in systems presenting a cyclic execution pattern, that is, repetition over time, making the future behavior of the tasks predictable. An example of this is Parallel Discrete Event Simulation (PDES), where several tasks are repeatedly executed in response to certain events. A comparison between the proposed algorithm and other load-balancing algorithms found in the literature reveals consistently better execution times with improvements in both load-balancing and locality of references that can be of help on current multicore desktop computers.

This paper proposes a design and structure of script language which is intended for easy description and prototyping of high-level image processing operations. The image operations are meant to be composed from basic building blocks represented by either C/C++ functions or appropriate block connections in FPGA (Field-Programmable Gate Array) circuits. The proposed language is designed for use in systems for rapid prototyping and testing of image processing applications as well as for final implementations of the applications. The integration of language into such systems is discussed as well as explanations of parts of the image processing system as seen through the interface of the proposed scripting language. The paper targets structures and syntax of the language, parallelization of high-level image operations and communication between the multiple instances of interpreters of the scripts.

The paper examines the usage potential of n-tuple algebra (NTA) developed by the authors as a theoretical generalization of structures and methods applied in intelligence systems. NTA supports formalization of a wide set of logical problems (abductive and modified conclusions, modelling of graphs, semantic networks, expert rules, etc.). This article mostly describes implementation of logical inference by means of NTA. Logical inference procedures in NTA can include, besides the known logical calculus methods, new algebraic methods for checking correctness of a consequence or for finding corollaries to a given axiom system. Inference methods consider (above feasibility of certain substitutions) inner structure of knowledge to be processed, thus providing faster solving of standard logical analysis tasks. Matrix properties of NTA objects allow to decrease laboriousness of intellectual procedures as well as to efficiently parallel logical inference algorithms. In NTA, we discovered new structural and statistical classes of conjunctive normal forms whose satisfiability can be detected for polynomial time. Consequently, many algorithms whose complexity evaluation is theoretically high, e.g. exponential, can in practice be solved in polynomial time, on the average. As for making databases more intelligent, NTA can be considered an extension of relational algebra to knowledge processing. In the authors' opinion, NTA can become a methodological basis for creating knowledge processing languages.

A set of n times n Boolean matrices along with the Boolean matrix multiplication operation form a semigroup. For each matrix A it is possible to find index mu and period łambda, such that A mu = A mu + łambda, and mu, łambda are the smallest positive integers with this property. We are concerned with a question: How many n times n Boolean matrices have the given index, and period? A new algorithm is presented that was used to compute index and period statistics of all square Boolean matrices up to n=8. Computed statistics are presented in the appendix of the paper.

The distributed computing environments, e.g. clouds, often deal with huge amounts of data, which constantly increase. The global growth of data is caused by ubiquitous personal devices, enterprise and scientific applications, etc. As the size of data grows new challenges are emerging in the context of storage management. Modern data and storage resource management systems need to face wide range of problems – minimizing energy consumption (green data centers), optimizing resource usage, throughput and capacity, data availability, security and legal issues, scalability. In addition users or their applications can have QoS (Quality of Service) requirements concerning the storage access, which further complicates the management. To cope with this problem a common mass storage system model taking into account the performance aspects of a storage system becomes a necessity. The model described with semantic technologies brings a semantic interoperability between the system components. In this paper we describe our approach at data management with QoS based on the developed models as a case study for distributed environments.