Derivative free optimization algorithms are implementations of trust region based derivative-free methods using multivariate polynomial interpolation. These are designed to minimize smooth functions whose derivatives are not available or costly to compute. The trust region based multilevel optimization algorithms for solving large scale unconstrained optimization problems resulting by discretization of partial differential equations (PDEs), make use of different discretization levels to reduce the computational cost. In this thesis, a derivative free multilevel optimization algorithm is derived and its convergence behavior is analyzed. The effectiveness of the algorithms is demonstrated on a shape optimization problem.

In this study, a new selective and time-window routing problem is defined for the first time in the literature, which is called the campaign routing problem (CRP). The two special cases of the CRP correspond to the two real-life problems, namely political campaign routing problem (PCRP) and the experiments on wheels routing problem (EWRP). The PCRP is based on two main decision levels. In the first level, a set of campaign regions is selected according to a given criteria subject to the special time-window constraints. In the second level, a pair of selected regions or a single region is assigned to a campaign day. In the EWRP, a single selected region (school) is assigned to a campaign day. These two problems are modeled using classical mathematical programming and bi-level programming methods, and a two-step heuristic approach is developed for the solution of the problems. Implementation of the solution methods is done using the test instances that are compiled from the real-life data. Computational results show that the solution methods developed generate good solutions in reasonable time.

Image annotation is defined as generating a set of textual words for a given image, learning from the available training data consisting of visual image content and annotation words. Methods developed for image annotation usually make use of region clustering algorithms to quantize the visual information. Visual codebooks are generated from the region clusters of low level visual features. These codebooks are then, matched with the words of the text document related to the image, in various ways. In this thesis, we propose a new image annotation technique, which improves the representation and quantization of the visual information by employing the available but unused information, called side information, which is hidden in the system. This side information is used to semi-supervise the clustering process which creates the visterms. The selection of side information depends on the visual image content, the annotation words and the relationship between them. Although there may be many different ways of defining and selecting side information, in this thesis, three types of side information are proposed. The first one is the hidden topic probability information obtained automatically from the text document associated with the image. The second one is the orientation and the third one is the color information around interest points that correspond to critical locations in the image. The side information provides a set of constraints in a semi-supervised K-means region clustering algorithm. Consequently, in generation of the visual terms from the regions, not only low level features are clustered, but also side information is used to complement the visual information, called visterms. This complementary information is expected to close the semantic gap between the low level features extracted from each region and the high level textual information. Therefore, a better match between visual codebook and the annotation words is obtained. Moreover, a speedup is obtained in the modified K-means algorithm because of the constraints brought by the side information. The proposed algorithm is implemented in a high performance parallel computation environment.

In this thesis a natural language query interface is developed for semantic and spatio-temporal querying of MPEG-7 based domain ontologies. The underlying ontology is created by attaching domain ontologies to the core Rhizomik MPEG-7 ontology. The user can pose concept, complex concept (objects connected with an &ldquo / AND&rdquo / or &ldquo / OR&rdquo / connector), spatial (left, right . . . ), temporal (before, after, at least 10 minutes before, 5 minutes after . . . ), object trajectory and directional trajectory (east, west, southeast . . . , left, right, upwards . . . ) queries to the system. Furthermore, the system handles the negative meaning in the user input. When the user enters a natural language (NL) input, it is parsed with the link parser. According to query type, the objects, attributes, spatial relation, temporal relation, trajectory relation, time filter and time information are extracted from the parser output by using predefined rules. After the information extraction, SPARQL queries are generated, and executed against the ontology by using an RDF API. Results are retrieved and they are used to calculate spatial, temporal, and trajectory relations between objects. The results satisfying the required relations are displayed in a tabular format and user can navigate through the multimedia content.

Economic events of this era reflect the fact that the value of information and technology has surpassed the value of physical production. This motivates countries to focus on increasing the education levels of citizens. However, policy making about education system and its returns requires dynamical analyses in order to be sustainable. The study aims to investigate the dynamic characteristics of a country-wide education system, in particular, that of Turkey. System Dynamics modeling, which is one of the most commonly referred tools for understanding the complex social structures, is used. Our model introduces dynamic relationships among different classes of labor forces with varying education levels, university admissions, research quality, and the investments made in education, research and other sectors. Model experimentation provides new insights into the investment and capacity-related aspects of the education system environment.

Recently, there has been much interest in the resilience of complex networks to random failures and intentional attacks. The study of the network robustness is particularly important by several occasions. In one hand a higher degree of robustness to errors and attacks may be desired for maintaining the information flow in communication networks under attacks. On the other hand planning a very limited attack aimed at fragmenting a network by removal of minimum number of the most important nodes might have significant usage in drug design. Many real world networks were found to display scale free topology including WWW, the internet, social networks or regulatory gene and protein networks. In the recent studies it was shown that while these networks have a surprising error tolerance, their scale-free topology makes them fragile under intentional attack, leaving the scientists a challenge on how to improve the networks robustness against attacks. In this thesis, we studied the protein domain co-occurrence network of yeast which displays scale free topology generated with data from Biomart which links to Pfam database. Several networks obtained from protein domain co-occurrence network having exactly the same connectivity distribution were compared under attacks to investigate the assumption that the different networks with the same connectivity distribution do not need to have the same attack tolerances. In addition to this, we considered that the networks with the same connectivity distribution have higher attack tolerance as we organize the same resources in a better way. Then, we checked for the variations of attack tolerance of the networks with the same connectiviy distributions. Furthermore, we investigated whether there is an evolutionary mechanism for having networks with higher or lower attack tolerances for the same connectivity distribution. As a result of these investigations, the different networks with the same connectivity distribution do not have the same attack tolerances under attack. In addition to this, it was observed that the networks with the same connectivity distribution have higher attack tolerances as organizing the same resources in a better way which implies that there is an evolutionary mechanism for having networks with higher attack tolerance for the same connectivity distribution.

In this thesis, we present an event orderer application that works on Turkish texts. Events are words denoting an occurrence or happenings in natural language texts. By using the features of the events in a sentence or by the helps of temporal expressions in the sentence, anchoring an event on a timeline or ordering events between other events are called event ordering. The application presented in this thesis, is one of the earliest study in this domain with Turkish and it realizes all needed sub modules for event ordering. It realizes event recognition in Turkish texts and event feature detection in Turkish texts. In addition to this, the application is realizing temporal expression recognition and temporal signal recognition tasks.

Inadequate or weak preparatory work before starting construction of any structure may cause serious problems during the construction period. For example, projects without sufficient detailed drawings or construction schedules and a disorganized building site can create many problems in the management and completion of the construction works. Consequently, the cost of construction increases digressively, the construction duration of the project extends and the quality of construction is affected adversely. This study dwells on the importance of construction schedules in achieving the aim of producing good quality construction work within the specified duration. Monitoring continuously the interactive relation concerning delays in construction schedules and contractor demands is a complicated process. Here the simplest and basic approach is that, both for owner and contractor, time is money and for this reason construction schedule delays should be analyzed and corrective measures should be taken in a timely manner. The main purpose of this study is to investigate the causes of construction schedule delays and the methods of schedule delay analyses. In this context completion construction works of a covered swimming pool building in Ankara was selected as a case study for analyzing project scheduling and the delays therein. The &ldquo / Time Impact Analysis Method&rdquo / (TIA) was applied to the case study project using PRIMAVERA&reg / software in order to determine the construction schedule delays / to measure the impacts of these delays on the project completion duration / and to allocate responsibility amongst the project participants for preventing delay claims. After the application of the delay analysis it was observed that the delays in the critical activities extended the project duration by 57 days in total i.e. by 15.4 % of the estimated construction period. Fines should have been paid by the contractor because of 31-days non-excusable delays. Also, the contractor should have been given a time extension of 26-days due to 22-days excusable compensable delays and 4-days excusable non-compensable delays which were beyond the control of the contractor. These delays were caused due to organizational deficiencies of the owner, the bureaucracy of the provincial municipality, the lack of detail drawings during the municipality application, the lack of experience of the contractor, problems in material procurement, unforeseeable weather conditions and shortages of qualified employees of the subcontractors. It was observed that of these all except one correspond to the important causes of delays as reported in literature concerning public projects in Turkey.

A lot of research have been carried in the field of optical character recognition. Selection of a feature extraction scheme is probably the most important factor in achieving high recognition performance. Fourier and wavelet transforms are among the popular feature extraction techniques allowing rotation invariant recognition. The performance of a particular feature extraction technique depends on the used dataset and the classifier. Dierent feature types may need dierent types of classifiers. In this thesis Fourier and wavelet based features are compared in terms of classification accuracy. The influence of noise with dierent intensities is also analyzed. Character recognition system is implemented with Matlab. Isolated gray scale character image first transformed into one dimensional function. Then, set of features are extracted. The feature set are fed to a classifier. Two types of classifier were used, Nearest Neighbor and Linear Discriminant Function. The performance of each feature extraction and classification methods were tested on various rotated and scaled character images.

We propose an interactive preference-based evolutionary algorithm for the clustering problem. The problem is highly combinatorial and referred to as NP-Hard in the literature. The goal of the problem is putting similar items in the same cluster and dissimilar items into different clusters according to a certain similarity measure, while maintaining some internal objectives such as compactness, connectivity or spatial separation. However, using one of these objectives is often not sufficient to detect different underlying structures in different data sets with clusters having arbitrary shapes and density variations. Thus, the current trend in the clustering literature is growing into the use of multiple objectives as the inadequacy of using a single objective is understood better. The problem is also difficult because the optimal solution is not well defined. To the best of our knowledge, all the multiobjective evolutionary algorithms for the clustering problem try to generate the whole Pareto optimal set. This may not be very useful since majority of the solutions in this set may be uninteresting when presented to the decision maker. In this study, we incorporate the preferences of the decision maker into a well known multiobjective evolutionary algorithm, namely SPEA-2, in the optimization process using reference points and achievement scalarizing functions to find the target clusters.