Volume 3 - 2011
Analysis of ISA server characteristics
Current moment in time, characterized by the explosion of Internet com munications, with simultaneous increase of security risks, imposes the need to place a security product on the network edge. Microsoft Internet Security and Acceleration Server 2006 is such product. It features several key security functionalities, usability, interface simplici ty and scalability, especially if Enterprise flavor is used. This document in vestigates key features of aforementioned product, which are necessary for the successful deployment in real world situations.
Social graph clustering
In this paper, we will work on a problem of social graph clustering. Problem of graph clustering is very well studied, but in almost all cases, disjunctive and total clustering is created. In a social graph, it is obvious that we can have people that belong to many groups, as is the case for the majority of people. But we can also have a person that does not belong to any group. We devise few algorithms that can be used to solve this problem. Also, currently, there is no good general metric for measuring quality of such clustering, so we created one that best suits the needs of specified task.
Using parallel methods for computing Monte Carlo integration
Monte Carlo methods are used when theory is impossible to apply or when the theory doesn’t exist. On the other hand, parallelisation is a powerful tool for speeding up the execution of an algorithm. In this paper, it is analyzed how parallelisation of Monte Carlo methods, Quazi Monte Carlo (MISER) algorithms by using MPI library can speed up the calculation of one dimensional and multidimensional integrals.
Optimization algorithms for solving the traveling salesman problem and their parallelization
Traveling salesman problem is often used for testing and comparing the performance of diffrent optimization algorithms. Here we present two of the most common algorithms for solving the problem: genetic algorithm and ant colony optimization. Detailed explanation and implementation notes are given as well as the results of testing. Also, we present a way to parallelize both of these algorithms and speedup that is achieved.
Детекција и класификација саобраћајних знакова у видео режиму рада
У овом раду представљен је метод за детекцију и класификацију саобраћајних знакова на видео и записима дигиталне слике. Детекција саобраћајних знакова и њихова идентификација има неколико примена, али је мало радова који описују детаљно ове процесе. Информације о облику су коришћене за детекцију и изоловање саобраћајног знака, а информације о боји и особинама хистограма изолованог облика се користе за класификовање. Број могућих облика саобраћајних знакова је мали, па је релативно лако детектовати и изоловати саобраћајне знакове. Међутим, квалитет слике, варијације у осветљености или изобличења знакова отежавају процес исправне класификације знакова.
OBJECT ORIENTED REALIZATION OF PERCEPTUAL COMPUTER AND A NOVEL INFERENCE OPERATOR
In this paper are presented introductory considerations of object oriented realization of Mendel's perceptual computer. The practical side of such realization is that it provides ability to easy develop and test novel elements. Particularly, for this paper is interesting possibility to replace the concept of novel weighted averages with another inference operator. Novel concept of weighted geometric shapes has been presented and empirically compared to novel weighted averages. The well known problem of missile system selection, represented as a hierarchical decision making problem, has been used for comparison. The paper announces novel theoretical basis for validation and comparing the operators.
Domain-Independent and Unsupervised Named Entity Extraction Web- Based Systems
Semantic Web is the future of the current Web. The transition from the current Web to the Semantic Web involves named entitiy extraction from today's Web in order to create RDF statements. Since the amount of information on the Web grows exponentially, this task requires a powerful unsupervised, doman-independent entitiy extraction system. In this paper we present and compar five existing systems. Also, proposal for the new entity extraction system is given.