34
Archived Theses
0
DOIs Assigned
0%
DOI Rate
Archived Theses
A neural-statistical modeling approach for keystroke recognition algorithms
The main problem of the computer and information systems is the security, which is to protectthe system from the attacks of imposter or unauthorized users. In order to supply bettersecurity, it must be determined clearly while system access that if the claimed one isauthorized user known by the system or not.Recently, biometric security systems technology is developed and added to the typicalauthentication systems , which are consist of username and PIN or password query, aiming toget higher security in system access.The keystroke pattern recognition system is chosen as one of the biometric security systemand proposed to perform a classification in this thesis. In order to achieve this, a perspective isdeveloped under the knowledge of the classification algorithms used earlier in keystrokepattern recognition systems. According to this, a model is designed which uses hybridcombination of two different algorithms. One of them is the statistical algorithm which is thevery firstly used one in pattern recognition and the other one is the neural networks. In themodel, the statistical algorithm formulations are embedded into the neural networkarchitecture. Designed algorithm model is described in detail and tested with sample userdatasets and performance results are presented.When thinking about need of new approaches in the classification algorithms in keystrokepattern recognition, this study can be a starting point to further enhancements with itsperspective on the subject.
A rule based expert system generation framework using an open source business rule engine
Knowledge is key instrument for the deciding processes. On the other hand, for a deciding process, gathering knowledge and learning are very difficult phases. For this reason, in the last decades, studies are focused on the machine-learning systems and the expert systems for the most of the knowledge oriented areas, like academic, commercial, military and industrial areas. In this thesis, a framework is developed for the rule base learning expert systems. Briefly, this framework will take a data set, induct the rules from this data set, construct an expert system according to inducted rules, and give a web based interface for testing new cases. There are a lot of concepts in this study. Classification, decision tree, knowledge acquisition, ID3 algorithm, rule base systems, expert systems, rule engines, open source perspective are some of them. These concepts will be discussed briefly, after the discussion; framework will be explained with some examples. Examples will show the reusability of the framework. Different data set can be applied the framework. But data set must be convenient to the ID3 decision tree algorithm. Other restrictions will be defined next sections. After constructing expert system new cases can be tested. This framework has some principles: ? Java technologies are used ? Open source tools are used where needed ? Standardizations are applied where available As a result of these principles, usability of the framework is dramatically increased.
Service oriented architecture
I present a survey that describes service oriented architecture. I made an sample applicationon Oracle BPEL by using Oracle SOA Suite for demonstration issues using it. I maderesearches about properties of service oriented architecture, advantages and disadvantages ofit. Besides, the past studies that were made by using SOA has been analyzed. In this thesisprinciples of service oriented architecture are explained.Service oriented architecture can be used in different areas by the help of its properties andadvantages. It can be adopted to different platforms easily. The use of SOA in applicationsdecreases the development time but making services at the beginning is not good in termseffort. The main advantage of service oriented architecture is reusability, a service can beused in different applications and there is no need to make any changes on the service. Tohave the description of any service is adequate to use it. A service can be used by sending thenecessary parameters to the service.In application part, an application has been developed step by step by using the properties ofservice oriented architecture.
Real-time hybrid parallel rendering
In computer graphics, rendering is described as the process of converting a description of a scene to an image. When the scene is complex and high quality images are required, the rendering process becomes computationally demanding. To provide the satisfactory performance, real-time computing techniques must be developed. Although parallelism has been extensively used in computer graphics for a long time, its initial use was primarily in specialized applications. Today, parallel computing is used in commodity personal computers, and various software-based rendering systems have been developed for general purpose real-time systems.As the new GPUs released to the market, the available rendering performance increases constantly. Also more powerful multi-core CPUs that have enabled more flexible and faster software-based graphics, such as real-time ray tracing. Despite this tremendous hardware development progress in rendering power, there will always be some applications that require distributed configurations for rendering. In this thesis, I present a prototype solution consisting of a system that supports different rendering modules (e.g., rasterization, and ray tracing) and combine it with a distributed graphics processing.This thesis provides a general introduction to the subject of real-time rendering, covering both hardware and software aspects. The main focus is on the underlying concepts and the issues which arise in the design of real-time rendering algorithms and systems. Different types of parallelism and how they can be applied in rendering applications are examined. Concepts from parallel computing, such as data decomposition, task granularity, scalability, and load balancing, are considered in relation to the rendering problem. Also concepts from computer graphics, such as coherence, culling, and level of detail which have a significant impact on the structure of parallel rendering algorithms are explored.
A multithreaded web crawler and text search engine
Without a doubt, internet is one of the best inventions in the last era. Number of internet users is more than millions. When internet users need information about something or somewhere, they visit search web sites or personal blog pages on the internet. For this purpose, many internet applications have been developed.Search Engines and data mining have shown a big improvement in the last 20 years. The developments on the internet increased the need of accessing and finding correct web resources. Raise of search engines caused to differentiation of search engine services. More intelligent search engines are important for accessing to the correct data.Search engines scan contents of the web sites and create indexes for their contents into own database using robots. Advances in search engines enable classification of subjects of the documents besides words or terms used in a document. Such search engines which have document classification property are called ?Clustered Search Engines?. For determination of page categories, the data mining methods are used.In this thesis study, a web crawler and classification system has been developed. The Open Directory Project (DMOZ) is used as a training set for the classification system. The labeled (categorized) web pages which are stored in the DMOZ directory are used as an input for the classification algorithms. We used classification algorithms available in WEKA Data Mining Tool. The web crawler developed in this thesis classifies web pages according to their subjects while scanning the web pages.
Finding the best performing solution algorithm for QAP
The quadratic assignment problem (QAP) of NP-Hard problems class is known as one of the hardest combinatorial optimization problems. In this thesis, a search is performed on the metaheuristics that have recently found widespread application in order to identify a heuristic procedure that performs well with the QAP. Algorithms which reflect implementations of Simulated Annealing, Genetic Algorithm, Scatter Search and Grasp ? type metaheuristics are tested and using real test problems these algorithms are compared. Same set of algorithms are tested on general QAP problems and observation to identify successful algorithms is made. To conclude the best performing heuristic is not easy to name due to the fact that the performance of a heuristic depends on the context of the problem, which determines the structure and relationships of problem parameters.
Otomatik hiperlink üretimi
One of the most important inventions of today is the Internet. Hundreds of millions of people anytime, from anywhere, can enter the Internet. On the Internet web pages are represented in HTML format and pages are linked through hyperlinks. Normally hyperlinks are defined by users manually. The objective of this thesis is to design a system that generates hyperlinks automatically.For this objective, a robot called SeaGEN has been developed. The robot SeaGEN analyses a web page and generates hyperlinks for certain words/phrases.During this Master study data mining techniques were used to generate hyperlinks. A classification system was developed. A training dataset was collected from Vikipedi. This training set was used for training the classification system. WEKA open source data mining software program was used for classification. The trained classification system generates hyperlinks automatically for a given set of pages.
Electromagnetic scattered field analysis of 2D wedge geometries with HFA techniques and FDTD method
In this thesis, electromagnetic scattered field analysis of two dimensional non-penetrable wedge geometry have been investigated with both high frequency asymptotic techniques (HFA) and finite difference time domain (FDTD) method.Among HFA techniques, Physical optics (PO), Physical theory of diffraction (PTD), Unified theory of diffraction (UTD), Parabolic equation (PE), Exact series and Exact integral methods are applied to inspect electromagnetic scattering behavior of two dimensional non-penetrable wedge geometry analytically.As a numerical technique, finite difference time domain (FDTD) method and the important aspects in its implementation are explained briefly. Also, absorbing boundary conditions and modeling issues are investigated. Meshing algorithm is developed to reduce staircase modeling errors which is formed by the application of standard Yee algorithm. In order to show the qualification of FDTD method, some examples are presented with HFA results.A novel Matlab based softwares WEDGE GUI and WEDGE FDTD GUI are also presented with thesis. The former was developed to analyze two dimensional perfect electric conductor (PEC) wedge geometries with various HFA methods. The latter was developed to analyze same geometry with finite difference time domain (FDTD) method. Source codes of these programs can be found in the CD given with this thesis.
Investigations on the effects of fourwave-mixing in dwdm fiber channels
Optical fiber cables have become ideal for achieving the challenge of reaching high speed data rates and terabit transmission, due to its low attenuation characteristics and high bandwidth capability such as 50 THz. Multiplexing of numerous channels on the same fibre requires higher transmit power or sufficiently lower fibre losses to utilize the available bandwidth having high bit rate with a span of thousands of kilometer.It is required to increase the number of optical channels in dense WDM systems in order to meet the extreme capacity demands on data transmission networks. This necessity of increment in number of channels can be provided only with small channel spacing. In order to reach required amount of signal channels, new frequency standards such as 25 GHz and 12.5 GHz were recently specified by ITU. In such narrower channel spacing with large number of WDM channels, the non-linear effects of the optical fibre can induce serious system impairments. In modern WDM systems, the primary nonlinear effects are cross phase modulation (XPM), and the four-wave mixing (FWM). The FWM characteristics are especially related to frequency allocation of channels. Throughout this thesis, we will be mainly dealing with the FWM impairments.The occurrence of FWM depends on several factors, such as frequency spacing between channels, the input power per channel, the dispersion characteristics of the optical fiber, and the distance along which the channels interact.Four wave mixing results from changes in the refractive index with optical power called optical Kerr effect. In FWM, two co-propagating waves produce two new optical sideband waves at different frequencies. The generation of these beat signals which fall into original signal wavelengths causes a channel energy loss, which induces a crosstalk effect. In long haul transmission links, the deployment of optical amplifiers makes the problem even worse, as not only the transmitted signal is amplified, but also the generated FWM products, which mix again with the signals, causing new products. Also when intense incident signal power launched into a fiber, linearity of optical response is lost. Moreover, in the usage of dispersion-shifted fibers (DSF), the FWM mechanism is enhanced, due to a depletion of the phase mismatch associated to the fiber's chromatic dispersion. In consequence, the detected signal power will fluctuate considerably.Numerous techniques have been proposed to minimize the detrimental limiting effects of FWM such as: the spectral allocation of the channels, the spectral assignment of the channels as far as possible from the zero-dispersion wavelength (?ZD), and the spectral distribution of the unequally spaced channels, which requires a complex system design.In this study, the impact of the channel spacing (positioning of the DWDM channels), phase mismatching, changing of channel input power and fiber length on FWM efficiency were analyzed based on represented algorithm. For various types of fibers such as G.652 (Single- Mode Fiber - SMF), G.653 (Dispersion-Shifted Fiber - DSF), and G.655 (Non-Zero Dispersion-Shifted Fiber - NZDSF) compliant fibers, considering the DWDM grids suggested by the ITU-T Recommendations G.692, and G.694.1, with uniform channel spacing of 100, 50, 25, and 12.5 GHz were compared with simulations. Split Step Fourier Method (SSFM) numerical technique has been used to model nonlinear Schrödinger (NLS) equation in order to investigate pulse propagation in optical fibers.
Empirical comparison of naïve bayes event models and smoothing methods for text classification
Naïve Bayes is one of the most commonly used algorithms in text classification due to its easy implementation and low complexity. There are two commonly referred event models in Naïve Bayes for text categorization; multivariate Bernoulli and multinomial models. A very large number of studies choose multinomial model and Laplace smoothing just based on the assumption that it performs better than multivariate model under almost any conditions. This thesis aims to shed some light into this widely adopted assumption by empirically analyzing Naïve Bayes event models and smoothing methods from a different perspective. In order to clarify the difference between these event models of Naïve Bayes, their classification performance are compared on different languages –English and Turkish-datasets. Results of our extensive experiments demonstrate that superior performance of multinomial model does not observed all the time. On the other hand, multivariate Bernoulli model can perform well when combined with an appropriate smoothing method under different training data size conditions at any training set size.