A pivotal moment in machine learning research was the creation of an important new research area, known as Ensemble Learning. In this work, we argue that ensembles are a very general concept, and though they have been widely used, they can be applied in more situations than they have been to date. Rather than using them only to combine the output of an algorithm, we can apply them to decisions made inside the algorithm itself, during the learning step. We call this approach Inner Ensembles. The motivation to develop Inner Ensembles was the opportunity to produce models with the similar advantages as regular ensembles, accuracy and stability for example, plus additional advantages such as comprehensibility, simplicity, rapid classification and small memory footprint. The main contribution of this work is to demonstrate how broadly this idea can be applied, and highlight its potential impact on all types of algorithms. To support our claim, we first provide a general guideline for applying Inner Ensembles to different algorithms. Then, using this framework, we apply them to two categories of learning methods: supervised and un-supervised. For the former we chose Bayesian network, and for the latter K-Means clustering. Our results show that 1) the overall performance of Inner Ensembles is significantly better than the original methods, and 2) Inner Ensembles provide similar performance improvements as regular ensembles.

This diploma thesis concerns approaches of management in the area of music ensembles on nowaday Czech music scene. Theoretical part of the thesis focuses on the music market as a whole, actual economical importance of music in the world, roles and tasks of managers and marketing communication. In the practical part of the thesis are introduced three music ensembles which are part of Czech music scene and furthemore one music ensemble that is located in the USA. In the thesis are described main geographical differences between American And Czech ensembles. In order to accomplish this practical part of thesis author used informaition mainly gained from the results of qualitative research which evolved around deeply detailed interviews. Main goal of the thesis is to describe methodology for managing smaller music ensembles mostly in the area of music groups focusing on non mainstream music genres. Author furthemore generelizes processes and approaches regarding the main differences of four music ensembles and their genre unlikeness.

Noble gas partitioning between supercritical CO2-H2O phases can be used to monitor Carbon Capture and Storage (CCS) sites and their natural analogues. However, in order for viable application, noble gas partitioning within these environments must be well constrained. Present estimates of partition coefficient for these systems are taken from the low pressure pure noble gas-water experiments of Crovetto et al. and Smith (Crovetto et al., 1982; Smith, 1985). The effect a supercritical CO2 phase may have on noble gas partitioning is assumed negligible, although this has not been empirically verified. In this work this assumption of noble gas behaviour within a supercritical CO2-H2O binary phase system is evaluated using a combined approach of experiment and simulation. Using a specially commissioned high pressure system at the British Geological Survey paired CO2 and H2O samples were collected from noble gas-enriched systems at pressures and temperatures ranging between 90 – 140 bar and 323.15 – 373.15 K. These were analysed for their noble gas content using a quadrupole mass spectrometer system developed specifically for this project. By comparing the relative concentrations of noble gases in each phase partition coefficients were defined for the experimental conditions. These were compared to their low pressure analogues. At higher CO2 densities all noble gases expressed a significant deviation from predicted partition coefficients. At the highest density (656 kg/m3) helium values decreased by -54% (i.e. reduced solubility within CO2) while argon, krypton and xenon values increased by 76%, 106% and 291% respectively. These deviations are due to supercritical CO2 acting as a polar solvent, the solvation power of which increases as a function of density. Polarisation is induced in each noble gas within this solvent based on their respective polarisabilities. Hence xenon, krypton and argon become more easily solvated as a function of CO2 density while solvating helium becomes harder. These deviation trends are well described using a second order polynomial. This fit defines a deviation coefficient which can be used to adapt low pressure partition coefficients to allow accurate predictions of partitioning within highly dense CO2 phases. Concurrently a Gibbs Ensemble Monte Carlo (GEMC) molecular model was iteratively developed to reproduce noble gas behaviour within these experimental systems. By optimising noble gas-water interactions a pure noble gas-water system was constructed for each noble gas at low pressure which replicated published partition coefficients. These optimised interactions were subsequently applied to low pressure CO2-H2O systems where partition coefficients were derived by calculating excess chemical potentials of noble gases in each phase. Again a good agreement was observed with published values. When the model was applied to the experimental conditions however, a poor agreement with the experimental values was observed. Instead simulated values replicated the low pressure Crovetto et al. and Smith datasets (Crovetto et al., 1982; Smith, 1985). This was due to no CO2-noble gas polarisation terms being included in the current iteration of the model. By including this within the model in the future a full reconciliation between the datasets is expected.

The Master's thesis is focused on analyse of human ressource management in proffesional ballet ensembles in the Czech republic. It describes the dance art via personnel activities -- human ressource planning, selection process, motivation, organizing work, education, development and employee renumeration. It calls attention to physical strain of this profession, to health and other risks, especially to short term of the active dance career and it shows the possibilities after the career termination.

A monoaminoxidase (MAO) é uma enzima importante, que pode atuar como alvo terapêutico. Inibidores da MAO-A apresentam atividade no tratamento de distúrbios de humor, enquanto os inibidores seletivos da MAO-B tem uso, especialmente, no tratamento da Doença de Parkinson. O conhecimento das interações ENZIMA-INIBIDOR é importante no planejamento de fármacos. Nesse contexto, foram realizados estudos das enzimas MAO-A e MAO-B com diferentes ligantes, através da combinação das metodologias de docking, Dinâmica Molecular e Ensemble Docking. Foram escolhidos os ligantes derivados da 1,4-naftoquinona (1,4-NQ), lapachol, menadiona, norlapachol, A2, B2 e C2, os inibidores comerciais clorgilina (MAO-A) e selegilina (MAO-B) e os substratos naturais serotonina (MAO-A) e dopamina (MAO-B). Os resultados do docking mostraram interação de todos os ligantes com algum dos resíduos da "gaiola aromática" (FAD, Tyr407/Tyr444 para MAO-A, Tyr398/Tyr435 para MAO-B), uma importante região catalítica da MAO. Além disso, a seletividade observada experimentalmente da menadiona com a MAO-B também foi observada no docking. Através da DM, foi possível observar algumas diferenças conformacionais entre as estruturas da MAO-A e MAO-B, que podem explicar a seletividade entre as duas isoformas, como por exemplo, distâncias entre resíduos do sítio ativo e ligações de hidrogênio. A partir do Ensemble Docking, foi verificado que a conformação do receptor influencia significativamente o escore das interações ENZIMA+LIGANTE para ligantes volumosos. / Monoamine oxidase (MAO) is an important enzyme that acts as therapeutic target. MAO-A inhibitors show pharmacological activity in the treatment of mood disorders, whereas MAO-B inhibitors are used especially in treatment of Parkinson's Disease. Knowledge of enzyme-inhibitor interactions is important in drug design. Therefore, studies of MAO-A and MAO-B enzymes with different ligands were performed by combining docking, Molecular Dynamics and Ensemble Docking methodologies. Ligands derived from 1,4-naphthoquinone, lapachol, menadione, nor-lapachol, A2, B2, C2, commercial inhibitors clorgyline (MAO-A) and selegiline (MAO-B) and the natural substrates serotonin (MAO-A) and dopamine (MAO-B) were chosen. The docking results shows interactions of all ligands with some residue of the "aromatic cage” (FAD cofactor, Tyr407/Tyr444 for MAO-A and Tyr398/Tyr435 for MAO-B), an important catalytic region of MAO. Furthermore, the experimentally observed selectivity of menadione with MAO-B was also observed by Docking. In Molecular Dynamics results, conformational differences were observed between MAO-A and MAO-B structures, which could explain the selectivity observed between isoforms, e.g. distances between residues of the active site and hydrogen bonds. Ensemble Docking results shows that the conformation of the receptor significantly influence the score of ENZYME+LIGAND interactions for bulky ligands.

Algoritmos de aprendizado de máquina são frequentemente os mais indicados em uma grande variedade de aplicações de mineração dados. Entretanto, a maioria das pesquisas em aprendizado de máquina refere-se ao problema bem definido de encontrar um modelo (geralmente de classificação) de um conjunto de dados pequeno, relativamente bem preparado para o aprendizado, no formato atributo-valor, no qual os atributos foram previamente selecionados para facilitar o aprendizado. Além disso, o objetivo a ser alcançado é simples e bem definido (modelos de classificação precisos, no caso de problemas de classificação). Mineração de dados propicia novas direções para pesquisas em aprendizado de máquina e impõe novas necessidades para outras. Com a mineração de dados, algoritmos de aprendizado estão quebrando as restrições descritas anteriormente. Dessa maneira, a grande contribuição da área de aprendizado de máquina para a mineração de dados é retribuída pelo efeito inovador que a mineração de dados provoca em aprendizado de máquina. Nesta tese, exploramos alguns desses problemas que surgiram (ou reaparecem) com o uso de algoritmos de aprendizado de máquina para mineração de dados. Mais especificamente, nos concentramos seguintes problemas: Novas abordagens para a geração de regras. Dentro dessa categoria, propomos dois novos métodos para o aprendizado de regras. No primeiro, propomos um novo método para gerar regras de exceção a partir de regras gerais. No segundo, propomos um algoritmo para a seleção de regras denominado Roccer. Esse algoritmo é baseado na análise ROC. Regras provêm de um grande conjunto externo de regras e o algoritmo proposto seleciona regras baseado na região convexa do gráfico ROC. Proporção de exemplos entre as classes. Investigamos vários aspectos relacionados a esse tópico. Primeiramente, realizamos uma série de experimentos em conjuntos de dados artificiais com o objetivo de testar nossa hipótese de que o grau de sobreposição entre as classes é um fator complicante em conjuntos de dados muito desbalanceados. Também executamos uma extensa análise experimental com vários métodos (alguns deles propostos neste trabalho) para balancear artificialmente conjuntos de dados desbalanceados. Finalmente, investigamos o relacionamento entre classes desbalanceadas e pequenos disjuntos, e a influência da proporção de classes no processo de rotulação de exemplos no algoritmo de aprendizado de máquina semi-supervisionado Co-training. Novo método para a combinação de rankings. Propomos um novo método, chamado BordaRank, para construir ensembles de rankings baseado no método de votação borda count. BordaRank pode ser aplicado em qualquer problema de ordenação binária no qual vários rankings estejam disponíveis. Resultados experimentais mostram uma melhora no desempenho com relação aos rankings individuais, alem de um desempenho comparável com algoritmos mais sofisticados que utilizam a predição numérica, e não rankings, para a criação de ensembles para o problema de ordenação binária. / Machine learning algorithms are often the most appropriate algorithms for a great variety of data mining applications. However, most machine learning research to date has mainly dealt with the well-circumscribed problem of finding a model (generally a classifier) given a single, small and relatively clean dataset in the attribute-value form, where the attributes have previously been chosen to facilitate learning. Furthermore, the end-goal is simple and well-defined, such as accurate classifiers in the classification problem. Data mining opens up new directions for machine learning research, and lends new urgency to others. With data mining, machine learning is now removing each one of these constraints. Therefore, machine learning's many valuable contributions to data mining are reciprocated by the latter's invigorating effect on it. In this thesis, we explore this interaction by proposing new solutions to some problems due to the application of machine learning algorithms to data mining applications. More specifically, we contribute to the following problems. New approaches to rule learning. In this category, we propose two new methods for rule learning. In the first one, we propose a new method for finding exceptions to general rules. The second one is a rule selection algorithm based on the ROC graph. Rules come from an external larger set of rules and the algorithm performs a selection step based on the current convex hull in the ROC graph. Proportion of examples among classes. We investigated several aspects related to this issue. Firstly, we carried out a series of experiments on artificial data sets in order to verify our hypothesis that overlapping among classes is a complicating factor in highly skewed data sets. We also carried out a broadly experimental analysis with several methods (some of them proposed by us) that artificially balance skewed datasets. Our experiments show that, in general, over-sampling methods perform better than under-sampling methods. Finally, we investigated the relationship between class imbalance and small disjuncts, as well as the influence of the proportion of examples among classes in the process of labelling unlabelled cases in the semi-supervised learning algorithm Co-training. New method for combining rankings. We propose a new method called BordaRanking to construct ensembles of rankings based on borda count voting, which could be applied whenever only the rankings are available. Results show an improvement upon the base-rankings constructed by taking into account the ordering given by classifiers which output continuous-valued scores, as well as a comparable performance with the fusion of such scores.

A oferta de produtos,informação e serviços a partir de perfis de usuários tem tornado os sistemas de recomendação cada vez mais presentes na Web, aumentando a facilidade de escolha e de permanência dos usuários nestes sistemas. Entretanto, existem otimizações a serem feitas principalmente com relação à modelagem do perfil do usuário. Geralmente, suas preferências são modeladas de modo superficial, devido à escassez das informações coletadas,como notas ou comentários, ou devido a informações indutivas que estão suscetíveis a erros. Esta dissertação propõe uma ferramenta de recomendação baseado em interações multimodais, capaz de combinar informações de usuários processadas individualmente por algoritmos de recomendação tradicionais. Nesta ferramenta desenvolveram-se quatro técnicas de combinação afim fornecer aos sistemas de recomendação, subsídios para melhoria na qualidade das predições em diversos domínios. / Providing products, information and services from user profiles has made the recommendation systems to be increasingly present, increasing the ease of selection and retention of users in Webservices. However, there are optimizations to be made in these systems mainly with respect to modeling the user profile. Generally, the preferences are modeled superficially, due to the scarcity of information collected, as notes or comments, or because of inductive information that is susceptible to errors. This work proposes are commendation tool based on multimodal interactions that combines users\' interactions, wich are processed individually by traditional recommendation algorithms. In this tool developed four combination of techniques in order to provide recommendation systems subsidies to improve the quality of predictions.

A grande quantidade de dados gerada em diversas áreas do conhecimento cria a necessidade do desenvolvimento de técnicas de mineração de dados cada vez mais eficientes e eficazes. Técnicas de agrupamento têm sido utilizadas com sucesso em várias áreas, especialmente naquelas em que não há conhecimento prévio sobre a organização dos dados. Contudo, a utilização de diferentes algoritmos de agrupamento, ou variações de um mesmo algoritmo, pode gerar uma ampla variedade de resultados. Tamanha variedade cria a necessidade de métodos para avaliar e selecionar bons resultados. Uma forma de avaliar esses resultados consiste em utilizar índices de validação de agrupamentos. Entretanto, uma grande diversidade de índices de validação foi proposta na literatura, o que torna a escolha de um único índice de validação uma tarefa penosa caso os desempenhos dos índices comparados sejam desconhecidos para a classe de problemas de interesse. Com a finalidade de obter um consenso entre resultados, é possível combinar um conjunto de agrupamentos ou índices de validação em uma única solução final. Combinações de agrupamentos (clustering ensembles) foram bem sucedidas em obter soluções robustas a variações no cenário de aplicação, o que faz do uso de comitês de agrupamentos uma alternativa interessante para encontrar soluções de qualidade razoável, segundo diferentes índices de validação. Adicionalmente, utilizar uma combinação de índices de validação pode tornar a avaliação de agrupamentos mais completa, uma vez que uma maioria dos índices combinados pode compensar o fraco desempenho do restante. Em alguns casos, não é possível lidar com um único conjunto de dados centralizado, por razões físicas ou questões de privacidade, o que gera a necessidade de distribuir o processo de mineração. Combinações de agrupamentos também podem ser estendidas para problemas de agrupamento de dados distribuídos, uma vez que informações sobre os dados, oriundas de diferentes fontes, podem ser combinadas em uma única solução global. O principal objetivo desse trabalho consiste em investigar técnicas de combinação de agrupamentos e de índices de validação aplicadas na seleção de agrupamentos para combinação e na mineração distribuída de dados. Adicionalmente, algoritmos evolutivos de agrupamento são estudados com a finalidade de selecionar soluções de qualidade dentre os resultados obtidos. As técnicas desenvolvidas possuem complexidade computacional reduzida e escalabilidade, o que permite sua aplicação em grandes conjuntos de dados ou cenários em que os dados encontram-se distribuídos / The large amount of data resulting from different areas of knowledge creates the need for development of data mining techniques increasingly efficient and effective. Clustering techniques have been successfully applied to several areas, especially when there is no prior knowledge about the data organization. Nevertheless, the use of different clustering algorithms, or variations of the same algorithm, can generate a wide variety of results, what raises the need to create methods to assess and select good results. One way to evaluate these results consists on using cluster validation indexes. However, a wide variety of validation indexes was proposed in the literature, which can make choosing a single index challenging if the performance of the compared indexes is unknown for the application scenario. In order to obtain a consensus among different options, a set of clustering results or validation indexes can be combined into a single final solution. Clustering ensembles successfully obtained results robust to variations in the application scenario, which makes them an attractive alternative to find solutions of reasonable quality, according to different validation indexes. Moreover, using a combination of validation indexes can promote a more powerful evaluation, as the majority of the combined indexes can compensate the poor performance of individual indexes. In some cases, it is not possible to work with a single centralized data set, for physical reasons or privacy concerns, which creates the need to distribute the mining process. Clustering ensembles can be extended to distributed data mining problems, since information about the data from distributed sources can be combined into a single global solution. The main objective of this research resides in investigating combination techniques for validation indexes and clustering results applied to clustering ensemble selection and distributed clustering. Additionally, evolutionary clustering algorithms are studied to select quality solutions among the obtained results. The techniques developed have scalability and reduced computational complexity, allowing their usage in large data sets or scenarios with distributed data

I den här kvalitativa studien har jag tittat närmare på de låtval man gör inom ämnet ensemble på gymnasieskolan och Kulturskolan. Jag har med hjälp av praxisgemenskap som teori och semistrukturerade intervjuer som metod av tre gymnasielärare, två lärare på Kulturskolan och en lärare på ett högstadium med musikinriktning sökt svar på följande forskningsfrågor: I vilken utsträckning uppdaterar lärare sin ensemblerepertoar med samtida låtar? Vad anser ensemblelärare om samtida musik ur ett pedagogiskt perspektiv? Vilka faktorer påverkar vad som används som undervisningsmaterial på ensemble lektionerna? Det jag fann var att det inom gymnasieskolan var få gånger det användes samtida låtar men att det var mer blandat på kulturskolan. Detta trots att lärarna egentligen inte såg något pedagogiskt hinder för att använda det i sin undervisning. De faktorer man hänvisade till angående varför man inte oftare använde samtida låtar i sin undervisning var bland annat skollag, att det kräver mer förberedelser och att man som lärare är mer trygg när man använder äldre, mer välkända låtar.

A la Maison d’Arrêt de Fresnes, on fait pratiquer le karaté aux détenus. Si les bienfaits du sport en prison sont un principe admis, qu’en est-il des effets pédagogiques de la pratique d'un sport de combat en milieu carcéral ? L’analyse des conduites motrices agressives en karaté, vérifiera si cette pratique apporte une régulation des affects, en comparant 188 observations réalisées en milieu associatif et 77 en détention. Ces observations discriminent l’agressivité licite, sur une échelle de valeurs graduées de -2 à 2 (sur la base d'indicateurs comportementaux objectifs annonciateurs du degré d'agressivité). Bousculant des idées reçues, les résultats révèlent que les karatékas associatifs (KA) sont plus agressifs – sur le plan « praxique » (1,71, écart-type = 0,58) et « kinésique » (1,42, é-t = 0,81) - que les karatékas détenus (KD) – respectivement 0,86, é-t = 1,17 et 0,3, é-t = 1,08. De leur côté, les processus de civilité s’expriment sur des moyennes de -2,00 pour les KA contre -0,81 pour les KD. Le karaté peut-il alors avoir une place en prison comme activité sportive ? La réponse semble affirmative. Sous réserve de processus ré-éducatif global, cette activité ne pouvant cependant garantir par elle-même solutionner le problème de la réinsertion. / At the Detention center of Fresnes, some prisoners go in for karate. If the benefits of sport in prison are an accepted principle, what about educational effects of the practice of a combat sport in prison? The analysis of sports practice aggressiveness in karate will verify if this activity brings a regulation of affects, by comparing 188 observations realized in associative environment and 77 in detention. These observations discriminate licit aggressiveness, on a scale of gradual values from 2 to 2 (on the basis of warning objective behavioral indicators of the degree of aggressiveness). Pushing aside preconceived ideas, the results reveal that the associative karateka (AK) is more aggressive - on the plan "praxique" (1,71, standard deviation = 0,58) and "kinesique" (1,42, s-d = 0,81) - than the prisoner karateka (PK) - respectively 0,86, s-d = 1,17 and 0,3, s-d = 1,08. On their side, the processes of civility express themselves on averages of -2,00 for the AK against -0,81 for the PK. Can the karate then have a place in prison as sports activity? The answer seems affirmative. Subject to global rehabilitation process, this activity can not however guarantee by itself to resolve the problem of the reintegration.