Learning Preferences with Kernel-Based Methods

Learning of preference relations has recently received significant attention in machine learning community. It is closely related to the classification and regression analysis and can be reduced to these tasks. However, preference learning involves prediction of ordering of the data points rather than prediction of a single numerical value as in case of regression or a class label as in case of classification. Therefore, studying preference relations within a separate framework facilitates not only better theoretical understanding of the problem, but also motivates development of the efficient algorithms for the task. Preference learning has many applications in domains such as information retrieval, bioinformatics, natural language processing, etc. For example, algorithms that learn to rank are frequently used in search engines for ordering documents retrieved by the query. Preference learning methods have been also applied to collaborative filtering problems for predicting individual customer choices from the vast amount of user generated feedback. In this thesis we propose several algorithms for learning preference relations. These algorithms stem from well founded and robust class of regularized least-squares methods and have many attractive computational properties. In order to improve the performance of our methods, we introduce several non-linear kernel functions. Thus, contribution of this thesis is twofold: kernel functions for structured data that are used to take advantage of various non-vectorial data representations and the preference learning algorithms that are suitable for different tasks, namely efficient learning of preference relations, learning with large amount of training data, and semi-supervised preference learning. Proposed kernel-based algorithms and kernels are applied to the parse ranking task in natural language processing, document ranking in information retrieval, and remote homology detection in bioinformatics domain. Training of kernel-based ranking algorithms can be infeasible when the size of the training set is large. This problem is addressed by proposing a preference learning algorithm whose computation complexity scales linearly with the number of training data points. We also introduce sparse approximation of the algorithm that can be efficiently trained with large amount of data. For situations when small amount of labeled data but a large amount of unlabeled data is available, we propose a co-regularized preference learning algorithm. To conclude, the methods presented in this thesis address not only the problem of the efficient training of the algorithms but also fast regularization parameter selection, multiple output prediction, and cross-validation. Furthermore, proposed algorithms lead to notably better performance in many preference learning tasks considered.

The Connectivity of Musical Aptitude and Foreign Language Learning Skills: Neural and Behavioural Evidence

Given the structural and acoustical similarities between speech and music, and possible overlapping cerebral structures in speech and music processing, a possible relationship between musical aptitude and linguistic abilities, especially in terms of second language pronunciation skills, was investigated. Moreover, the laterality effect of the mother tongue was examined with both adults and children by means of dichotic listening scores. Finally, two event-related potential studies sought to reveal whether children with advanced second language pronunciation skills and higher general musical aptitude differed from children with less-advanced pronunciation skills and less musical aptitude in accuracy when preattentively processing mistuned triads and music / speech sound durations. The results showed a significant relationship between musical aptitude, English language pronunciation skills, chord discrimination ability, and sound-change-evoked brain activation in response to musical stimuli (durational differences and triad contrasts). Regular music practice may also have a modulatory effect on the brain’s linguistic organization and cause altered hemispheric functioning in those who have regularly practised music for years. Based on the present results, it is proposed that language skills, both in production and discrimination, are interconnected with perceptual musical skills.

Neural correlates of language learning in adults

The human language-learning ability persists throughout life, indicating considerable flexibility at the cognitive and neural level. This ability spans from expanding the vocabulary in the mother tongue to acquisition of a new language with its lexicon and grammar. The present thesis consists of five studies that tap both of these aspects of adult language learning by using magnetoencephalography (MEG) and functional magnetic resonance imaging (fMRI) during language processing and language learning tasks. The thesis shows that learning novel phonological word forms, either in the native tongue or when exposed to a foreign phonology, activates the brain in similar ways. The results also show that novel native words readily become integrated in the mental lexicon. Several studies in the thesis highlight the left temporal cortex as an important brain region in learning and accessing phonological forms. Incidental learning of foreign phonological word forms was reflected in functionally distinct temporal lobe areas that, respectively, reflected short-term memory processes and more stable learning that persisted to the next day. In a study where explicitly trained items were tracked for ten months, it was found that enhanced naming-related temporal and frontal activation one week after learning was predictive of good long-term memory. The results suggest that memory maintenance is an active process that depends on mechanisms of reconsolidation, and that these process vary considerably between individuals. The thesis put special emphasis on studying language learning in the context of language production. The neural foundation of language production has been studied considerably less than that of perceptive language, especially on the sentence level. A well-known paradigm in language production studies is picture naming, also used as a clinical tool in neuropsychology. This thesis shows that accessing the meaning and phonological form of a depicted object are subserved by different neural implementations. Moreover, a comparison between action and object naming from identical images indicated that the grammatical class of the retrieved word (verb, noun) is less important than the visual content of the image. In the present thesis, the picture naming was further modified into a novel paradigm in order to probe sentence-level speech production in a newly learned miniature language. Neural activity related to grammatical processing did not differ between the novel language and the mother tongue, but stronger neural activation for the novel language was observed during the planning of the upcoming output, likely related to more demanding lexical retrieval and short-term memory. In sum, the thesis aimed at examining language learning by combining different linguistic domains, such as phonology, semantics, and grammar, in a dynamic description of language processing in the human brain.

Identifying first graders at risk of reading and writing difficulties : creating a group-based screening tool in kiswahili in Tanzania

Early identification of beginning readers at risk of developing reading and writing difficulties plays an important role in the prevention and provision of appropriate intervention. In Tanzania, as in other countries, there are children in schools who are at risk of developing reading and writing difficulties. Many of these children complete school without being identified and without proper and relevant support. The main language in Tanzania is Kiswahili, a transparent language. Contextually relevant, reliable and valid instruments of identification are needed in Tanzanian schools. This study aimed at the construction and validation of a group-based screening instrument in the Kiswahili language for identifying beginning readers at risk of reading and writing difficulties. In studying the function of the test there was special interest in analyzing the explanatory power of certain contextual factors related to the home and school. Halfway through grade one, 337 children from four purposively selected primary schools in Morogoro municipality were screened with a group test consisting of 7 subscales measuring phonological awareness, word and letter knowledge and spelling. A questionnaire about background factors and the home and school environments related to literacy was also used. The schools were chosen based on performance status (i.e. high, good, average and low performing schools) in order to include variation. For validation, 64 children were chosen from the original sample to take an individual test measuring nonsense word reading, word reading, actual text reading, one-minute reading and writing. School marks from grade one and a follow-up test half way through grade two were also used for validation. The correlations between the results from the group test and the three measures used for validation were very high (.83-.95). Content validity of the group test was established by using items drawn from authorized text books for reading in grade one. Construct validity was analyzed through item analysis and principal component analysis. The difficulty level of most items in both the group test and the follow-up test was good. The items also discriminated well. Principal component analysis revealed one powerful latent dimension (initial literacy factor), accounting for 93% of the variance. This implies that it could be possible to use any set of the subtests of the group test for screening and prediction. The K-Means cluster analysis revealed four clusters: at-risk children, strugglers, readers and good readers. The main concern in this study was with the groups of at-risk children (24%) and strugglers (22%), who need the most assistance. The predictive validity of the group test was analyzed by correlating the measures from the two school years and by cross tabulating grade one and grade two clusters. All the correlations were positive and very high, and 94% of the at-risk children in grade two were already identified in the group test in grade one. The explanatory power of some of the home and school factors was very strong. The number of books at home accounted for 38% of the variance in reading and writing ability measured by the group test. Parents´ reading ability and the support children received at home for schoolwork were also influential factors. Among the studied school factors school attendance had the strongest explanatory power, accounting for 21% of the variance in reading and writing ability. Having been in nursery school was also of importance. Based on the findings in the study a short version of the group test was created. It is suggested for use in the screening processes in grade one aiming at identifying children at risk of reading and writing difficulties in the Tanzanian context. Suggestions for further research as well as for actions for improving the literacy skills of Tanzanian children are presented.

Kernel-Based Ranking. Methods for Learning and Performance Estimation

Machine learning provides tools for automated construction of predictive models in data intensive areas of engineering and science. The family of regularized kernel methods have in the recent years become one of the mainstream approaches to machine learning, due to a number of advantages the methods share. The approach provides theoretically well-founded solutions to the problems of under- and overfitting, allows learning from structured data, and has been empirically demonstrated to yield high predictive performance on a wide range of application domains. Historically, the problems of classification and regression have gained the majority of attention in the field. In this thesis we focus on another type of learning problem, that of learning to rank. In learning to rank, the aim is from a set of past observations to learn a ranking function that can order new objects according to how well they match some underlying criterion of goodness. As an important special case of the setting, we can recover the bipartite ranking problem, corresponding to maximizing the area under the ROC curve (AUC) in binary classification. Ranking applications appear in a large variety of settings, examples encountered in this thesis include document retrieval in web search, recommender systems, information extraction and automated parsing of natural language. We consider the pairwise approach to learning to rank, where ranking models are learned by minimizing the expected probability of ranking any two randomly drawn test examples incorrectly. The development of computationally efficient kernel methods, based on this approach, has in the past proven to be challenging. Moreover, it is not clear what techniques for estimating the predictive performance of learned models are the most reliable in the ranking setting, and how the techniques can be implemented efficiently. The contributions of this thesis are as follows. First, we develop RankRLS, a computationally efficient kernel method for learning to rank, that is based on minimizing a regularized pairwise least-squares loss. In addition to training methods, we introduce a variety of algorithms for tasks such as model selection, multi-output learning, and cross-validation, based on computational shortcuts from matrix algebra. Second, we improve the fastest known training method for the linear version of the RankSVM algorithm, which is one of the most well established methods for learning to rank. Third, we study the combination of the empirical kernel map and reduced set approximation, which allows the large-scale training of kernel machines using linear solvers, and propose computationally efficient solutions to cross-validation when using the approach. Next, we explore the problem of reliable cross-validation when using AUC as a performance criterion, through an extensive simulation study. We demonstrate that the proposed leave-pair-out cross-validation approach leads to more reliable performance estimation than commonly used alternative approaches. Finally, we present a case study on applying machine learning to information extraction from biomedical literature, which combines several of the approaches considered in the thesis. The thesis is divided into two parts. Part I provides the background for the research work and summarizes the most central results, Part II consists of the five original research articles that are the main contribution of this thesis.

Uusien uhkakuvien luominen: tapaus 'kiinalaiset kybersoturit' : lingvistinen uhka-analyysi kyberdiskurssista

Since his inauguration, President Barack Obama has emphasized the need for a new cybersecurity policy, pledging to make it a "national security priority". This is a significant change in security discourse after an eight-year war on terror – a term Obama announced to be no longer in use. After several white papers, reports and the release of the so-called 60-day Cybersecurity Review, Obama announced the creation of a "cyber czar" position and a new military cyber command to coordinate American cyber defence and warfare. China, as an alleged cyber rival, has played an important role in the discourse that introduced the need for the new office and the proposals for changes in legislation. Research conducted before this study suggest the dominance of state-centric enemy descriptions paused briefly after 9/11, but returned soon into threat discourse. The focus on China's cyber activities fits this trend. The aim of this study is to analyze the type of modern threat scenarios through a linguistic case study on the reporting on Chinese hackers. The methodology of this threat analysis is based on the systemic functional language theory, and realizes as an analysis of action and being descriptions (verbs) used by the American authorities. The main sources of data include the Cybersecurity Act 2009, Securing Cyberspace for the 44th Presidency, and 2008 Report to Congress of the U.S. - China Economic and Security Review Commission. Contrary to the prevailing and popularized terrorism discourse, the results show the comeback of Cold War rhetoric as well as the establishment of a state-centric threat perception in cyber discourse. Cyber adversaries are referred to with descriptions of capacity, technological superiority and untrustworthiness, whereas the ‘self’ is described as vulnerable and weak. The threat of cyber attacks is compared to physical attacks on critical military and civilian infrastructure. The authorities and the media form a cycle, in which both sides quote each other and foster each other’s distrust and rhetoric. The white papers present China's cyber army as an existential threat. This leads to cyber discourse turning into a school-book example of a securitization process. The need for security demands action descriptions, which makes new rules and regulations acceptable. Cyber discourse has motives and agendas that are separate from real security discourse: the arms race of the 21st century is about unmanned war.

Software Security Vulnerabilities in Mobile Peer-to-peer Environment

Increase of computational power and emergence of new computer technologies led to popularity of local communications between personal trusted devices. By-turn, it led to emergence of security problems related to user data utilized in such communications. One of the main aspects of the data security assurance is security of software operating on mobile devices. The aim of this work was to analyze security threats to PeerHood, software intended for performing personal communications between mobile devices regardless of underlying network technologies. To reach this goal, risk-based software security testing was performed. The results of the testing showed that the project has several security vulnerabilities. So PeerHood cannot be considered as a secure software. The analysis made in the work is the first step towards the further implementation of PeerHood security mechanisms, as well as taking into account security in the development process of this project.

Enhancing human security through crisis management : opportunities and challenges for learning

Inhimilliseen turvallisuuteen kriisinhallinnan kautta – oppimisen mahdollisuuksia ja haasteita Kylmän sodan jälkeen aseelliset konfliktit ovat yleensä alkaneet niin sanotuissa hauraissa valtioissa ja köyhissä maissa, ne ovat olleet valtioiden sisäisiä ja niihin on osallistunut ei-valtiollisia aseellisia ryhmittymiä. Usein ne johtavat konfliktikierteeseen, jossa sota ja vakaammat olot vaihtelevat. Koska kuolleisuus konflikteissa voi jäädä alle kansainvälisen määritelmän (1000 kuollutta vuodessa), kutsun tällaisia konflikteja ”uusiksi konflikteiksi”. Kansainvälinen yhteisö on pyrkinyt kehittämään kriisinhallinnan ja rauhanrakentamisen malleja, jotta pysyvä rauhantila saataisiin aikaiseksi. Inhimillinen turvallisuus perustuu näkemykseen, jossa kunnioitetaan jokaisen yksilön ihmisoikeuksia ja jolla on vaikutusta myös kriisinhallinnan ja rauhanrakentamisen toteuttamiseen. Tutkimukseen kuuluu kaksi empiiristä osaa: Delfoi tulevaisuuspaneeliprosessin sekä kriisinhallintahenkilöstön haastattelut. Viisitoista eri alojen kriisinhallinta-asiantuntijaa osallistui paneeliin, joka toteutettiin vuonna 2008. Paneelin tulosten mukaan tulevat konfliktit usein ovat uusien konfliktien kaltaisia. Lisäksi kriisinhallintahenkilöstöltä edellytetään vuorovaikutus- ja kommunikaatiokykyä ja luonnollisesti myös varsinaisia ammatillisia valmiuksia. Tulevaisuuspaneeli korosti vuorovaikutus- ja kommunikaatiotaitoja erityisesti siviilikriisinhallintahenkilöstön kompetensseissa, mutta samat taidot painottuivat sotilaallisen kriisinhallinnan henkilöstön kompetensseissakin. Kriisinhallinnassa tarvitaan myös selvää työnjakoa eri toimijoiden kesken. Kosovossa työskennelleen henkilöstön haastatteluaineisto koostui yhteensä 27 teemahaastattelusta. Haastateltavista 9 oli ammattiupseeria, 10 reservistä rekrytoitua rauhanturvaajaa ja 8 siviilikriisinhallinnassa työskennellyttä henkilöä. Haastattelut toteutettiin helmi- ja kesäkuun välisenä aikana vuonna 2008. Haastattelutuloksissa korostui vuorovaikutus- ja kommunikaatiotaitojen merkitys, sillä monissa käytännön tilanteissa haastateltavat olivat ratkoneet ongelmia yhteistyössä muun kriisinhallintahenkilöstön tai paikallisten asukkaiden kanssa. Kriisinhallinnassa toteutui oppimisprosesseja, jotka usein olivat luonteeltaan myönteisiä ja informaalisia. Tällaisten onnistumisten vaikutus yksilön minäkuvaan oli myönteinen. Tällaisia prosesseja voidaan kuvata ”itseä koskeviksi oivalluksiksi”. Kriisinhallintatehtävissä oppimisella on erityinen merkitys, jos halutaan kehittää toimintoja inhimillisen turvallisuuden edistämiseksi. Siksi on tärkeää, että kriisinhallintakoulutusta ja kriisinhallintatyössä oppimista kehitetään ottamaan huomioon oppimisen eri tasot ja ulottuvuudet sekä niiden merkitys. Informaaliset oppimisen muodot olisi otettava paremmin huomioon kriisinhallintakoulutusta ja kriisinhallintatehtävissä oppimista kehitettäessä. Palautejärjestelmää olisi kehitettävä eri tavoin. Koko kriisinhallintaoperaation on saatava tarvittaessa myös kriittistä palautetta onnistumisista ja epäonnistumisista. Monet kriisinhallinnassa työskennelleet kaipaavat kunnollista palautetta työrupeamastaan. Liian rutiininomaiseksi koettu palaute ei edistä yksilön oppimista. Spontaanisti monet haastatellut pitivät tärkeänä, että kriisinhallinnassa työskennelleillä olisi mahdollisuus debriefing- tyyppiseen kotiinpaluukeskusteluun. Pelkkä tällainen mahdollisuus ilmeisesti voisi olla monelle myönteinen uutinen, vaikka tilaisuutta ei hyödynnettäisikään. Paluu kriisinhallintatehtävistä Suomeen on monelle haasteellisempaa kuin näissä tehtävissä työskentelyn aloittaminen ulkomailla. Tutkimuksen tulokset kannustavat tutkimaan kriisinhallintaa oppimisen näkökulmasta. On myös olennaista, että kriisinhallinnan palautejärjestelmiä kehitetään mahdollisimman hyvin edistämään sekä yksilöllistä että organisatorista oppimista kriisinhallinnassa. Kriisinhallintaoperaatio on oppimisympäristö. Kriisinhallintahenkilöstön kommunikaatio- ja vuorovaikutustaitojen kehittäminen on olennaista tavoiteltaessa kestävää rauhanprosessia, jossa konfliktialueen asukkaatkin ovat mukana.

Building object-oriented software with the D-Bus messaging system

Object-oriented programming is a widely adopted paradigm for desktop software development. This paradigm partitions software into separate entities, objects, which consist of data and related procedures used to modify and inspect it. The paradigm has evolved during the last few decades to emphasize decoupling between object implementations, via means such as explicit interface inheritance and event-based implicit invocation. Inter-process communication (IPC) technologies allow applications to interact with each other. This enables making software distributed across multiple processes, resulting in a modular architecture with benefits in resource sharing, robustness, code reuse and security. The support for object-oriented programming concepts varies between IPC systems. This thesis is focused on the D-Bus system, which has recently gained a lot of users, but is still scantily researched. D-Bus has support for asynchronous remote procedure calls with return values and a content-based publish/subscribe event delivery mechanism. In this thesis, several patterns for method invocation in D-Bus and similar systems are compared. The patterns that simulate synchronous local calls are shown to be dangerous. Later, we present a state-caching proxy construct, which avoids the complexity of properly asynchronous calls for object inspection. The proxy and certain supplementary constructs are presented conceptually as generic object-oriented design patterns. The e ect of these patterns on non-functional qualities of software, such as complexity, performance and power consumption, is reasoned about based on the properties of the D-Bus system. The use of the patterns reduces complexity, but maintains the other qualities at a good level. Finally, we present currently existing means of specifying D-Bus object interfaces for the purposes of code and documentation generation. The interface description language used by the Telepathy modular IM/VoIP framework is found to be an useful extension of the basic D-Bus introspection format.

Aging based maintenance and reinvestment scheduling of electric distribution

The maintenance of electric distribution network is a topical question for distribution system operators because of increasing significance of failure costs. In this dissertation the maintenance practices of the distribution system operators are analyzed and a theory for scheduling maintenance activities and reinvestment of distribution components is created. The scheduling is based on the deterioration of components and the increasing failure rates due to aging. The dynamic programming algorithm is used as a solving method to maintenance problem which is caused by the increasing failure rates of the network. The other impacts of network maintenance like environmental and regulation reasons are not included to the scope of this thesis. Further the tree trimming of the corridors and the major disturbance of the network are not included to the problem optimized in this thesis. For optimizing, four dynamic programming models are presented and the models are tested. Programming is made in VBA-language to the computer. For testing two different kinds of test networks are used. Because electric distribution system operators want to operate with bigger component groups, optimal timing for component groups is also analyzed. A maintenance software package is created to apply the presented theories in practice. An overview of the program is presented.

Scheduling of guarded command based models

Formal methods provide a means of reasoning about computer programs in order to prove correctness criteria. One subtype of formal methods is based on the weakest precondition predicate transformer semantics and uses guarded commands as the basic modelling construct. Examples of such formalisms are Action Systems and Event-B. Guarded commands can intuitively be understood as actions that may be triggered when an associated guard condition holds. Guarded commands whose guards hold are nondeterministically chosen for execution, but no further control flow is present by default. Such a modelling approach is convenient for proving correctness, and the Refinement Calculus allows for a stepwise development method. It also has a parallel interpretation facilitating development of concurrent software, and it is suitable for describing event-driven scenarios. However, for many application areas, the execution paradigm traditionally used comprises more explicit control flow, which constitutes an obstacle for using the above mentioned formal methods. In this thesis, we study how guarded command based modelling approaches can be conveniently and efficiently scheduled in different scenarios. We first focus on the modelling of trust for transactions in a social networking setting. Due to the event-based nature of the scenario, the use of guarded commands turns out to be relatively straightforward. We continue by studying modelling of concurrent software, with particular focus on compute-intensive scenarios. We go from theoretical considerations to the feasibility of implementation by evaluating the performance and scalability of executing a case study model in parallel using automatic scheduling performed by a dedicated scheduler. Finally, we propose a more explicit and non-centralised approach in which the flow of each task is controlled by a schedule of its own. The schedules are expressed in a dedicated scheduling language, and patterns assist the developer in proving correctness of the scheduled model with respect to the original one.

A bridge between engineering and language learning : automation and German in an online course

Artikkeli luettavissa osassa: Part 2. - ISBN 9789522163172(PDF). - Liitteenä työpaperi

Training executive functions in the laboratory and in real life : cognitive consequences of computer-based exercises and bilingualism

The question of the trainability of executive functions and the impact of such training on related cognitive skills has stirred considerable research interest. Despite a number of studies investigating this, the question has not yet been solved. The general aim of this thesis was to investigate two very different types of training of executive functions: laboratory-based computerized training (Studies I-III) and realworld training through bilingualism (Studies IV-V). Bilingualism as a kind of training of executive functions is based on the idea that managing two languages requires executive resources, and previous studies have suggested a bilingual advantage in executive functions. Three executive functions were studied in the present thesis: updating of working memory (WM) contents, inhibition of irrelevant information, and shifting between tasks and mental sets. Studies I-III investigated the effects of computer-based training of WM updating (Study I), inhibition (Study II), and set shifting (Study III) in healthy young adults. All studies showed increased performance on the trained task. More importantly, improvement on an untrained task tapping the trained executive function (near transfer) was seen in Study I and II. None of the three studies showed improvement on untrained tasks tapping some other cognitive function (far transfer) as a result of training. Study I also used PET to investigate the effects of WM updating training on a neurotransmitter closely linked to WM, namely dopamine. The PET results revealed increased striatal dopamine release during WM updating performance as a result of training. Study IV investigated the ability to inhibit task-irrelevant stimuli in bilinguals and monolinguals by using a dichotic listening task. The results showed that the bilinguals exceeded the monolinguals in inhibiting task-irrelevant information. Study V introduced a new, complementary research approach to study the bilingual executive advantage and its underlying mechanisms. To circumvent the methodological problems related to natural groups design, this approach focuses only on bilinguals and examines whether individual differences in bilingual behavior correlate with executive task performances. Using measures that tap the three above-entioned executive functions, the results suggested that more frequent language switching was associated with better set shifting skills, and earlier acquisition of the second language was related to better inhibition skills. In conclusion, the present behavioral results showed that computer-based training of executive functions can improve performance on the trained task and on closely related tasks, but does not yield a more general improvement of cognitive skills. Moreover, the functional neuroimaging results reveal that WM training modulates striatal dopaminergic function, speaking for training-induced neural plasticity in this important neurotransmitter system. With regard to bilingualism, the results provide further support to the idea that bilingualism can enhance executive functions. In addition, the new complementary research approach proposed here provides some clues as to which aspects of everyday bilingual behavior may be related to the advantage in executive functions in bilingual individuals.

A cyber security architecture for military networks using a cognitive network approach

Cyber security is one of the main topics that are discussed around the world today. The threat is real, and it is unlikely to diminish. People, business, governments, and even armed forces are networked in a way or another. Thus, the cyber threat is also facing military networking. On the other hand, the concept of Network Centric Warfare sets high requirements for military tactical data communications and security. A challenging networking environment and cyber threats force us to consider new approaches to build security on the military communication systems. The purpose of this thesis is to develop a cyber security architecture for military networks, and to evaluate the designed architecture. The architecture is described as a technical functionality. As a new approach, the thesis introduces Cognitive Networks (CN) which are a theoretical concept to build more intelligent, dynamic and even secure communication networks. The cognitive networks are capable of observe the networking environment, make decisions for optimal performance and adapt its system parameter according to the decisions. As a result, the thesis presents a five-layer cyber security architecture that consists of security elements controlled by a cognitive process. The proposed architecture includes the infrastructure, services and application layers that are managed and controlled by the cognitive and management layers. The architecture defines the tasks of the security elements at a functional level without introducing any new protocols or algorithms. For evaluating two separated method were used. The first method is based on the SABSA framework that uses a layered approach to analyze overall security of an organization. The second method was a scenario based method in which a risk severity level is calculated. The evaluation results show that the proposed architecture fulfills the security requirements at least at a high level. However, the evaluation of the proposed architecture proved to be very challenging. Thus, the evaluation results must be considered very critically. The thesis proves the cognitive networks are a promising approach, and they provide lots of benefits when designing a cyber security architecture for the tactical military networks. However, many implementation problems exist, and several details must be considered and studied during the future work.

Consistency of UML based designs using ontology reasoners

Software plays an important role in our society and economy. Software development is an intricate process, and it comprises many different tasks: gathering requirements, designing new solutions that fulfill these requirements, as well as implementing these designs using a programming language into a working system. As a consequence, the development of high quality software is a core problem in software engineering. This thesis focuses on the validation of software designs. The issue of the analysis of designs is of great importance, since errors originating from designs may appear in the final system. It is considered economical to rectify the problems as early in the software development process as possible. Practitioners often create and visualize designs using modeling languages, one of the more popular being the Uni ed Modeling Language (UML). The analysis of the designs can be done manually, but in case of large systems, the need of mechanisms that automatically analyze these designs arises. In this thesis, we propose an automatic approach to analyze UML based designs using logic reasoners. This approach firstly proposes the translations of the UML based designs into a language understandable by reasoners in the form of logic facts, and secondly shows how to use the logic reasoners to infer the logical consequences of these logic facts. We have implemented the proposed translations in the form of a tool that can be used with any standard compliant UML modeling tool. Moreover, we authenticate the proposed approach by automatically validating hundreds of UML based designs that consist of thousands of model elements available in an online model repository. The proposed approach is limited in scope, but is fully automatic and does not require any expertise of logic languages from the user. We exemplify the proposed approach with two applications, which include the validation of domain specific languages and the validation of web service interfaces.