Energy-Based Models (EBMs) are a family of unsupervised machine learning models that associate each point in the input space with an energy value in which low energy indicates a high likelihood. Specifically, an EBM can be viewed as an unnormalized probabilistic model, which upon normalization, gives rise to the probability density function of data. The main difficulty in learning EBMs lies in the computation of the normalization constant, or the partition function, a task known to be intractable in general. Several approaches have been proposed to avoid or overcome this difficulty, including Maximum Likelihood Estimation (MLE) with Markov chain Monte Carlo (MCMC), Score Matching (SM), Noise Contrastive Estimation (NCE), and so on. This thesis studies the learning of EBMs using NCE. Briefly, in NCE, the EBM learning problem is converted to learning a binary classifier, which aims to distinguish the real data from fake data drawn from a noise distribution. This process allows the learning of the energy function in the EBM to bypass a direct estimation of the partition function and a certain theoretical guarantee is available under some assumptions in some asymptotic limit. Despite the nice theoretical properties of NCE, in this work, we show that learning EBMs using NCE entails significant practical limitations. Specifically, there appears a tension between the quality of the learned model and the computational efficiency, due to which we must sacrifice one to achieve the other. We establish these limitations via empirical studies as well as a theoretical analysis based on a simple “Gaussian data learning problem”. Our analysis inspires a revised NCE scheme, Adaptive Noise Contrastive Estimation (ANCE), to overcome these limitations. Empirically, we show that ANCE achieves a better quality-efficiency trade-off than the standard NCE in some regimes.

A presente dissertação procura investigar que papel a Internet exerce nos projetos educomunicativos implementados pelo NCE Núcleo de Comunicação e Educação da Universidade de São Paulo, buscando entender como o Núcleo vem se apropriando da linguagem virtual nos diferentes programas implementados a partir de 2001, junto a redes públicas de educação. Para tanto, a pesquisa toma como amostragem os sites dos projetos Educom.TV (2002), TodeOlho.TV (2002), Educom.rádio (2001-2004) e Educomradio.centro-oeste (2005-2006). O trabalho busca identificar a especificidade do emprego da linguagem digital tanto no âmbito da formação presencial quanto no da formação a distância. Como metodologia, a pesquisa faz uso de entrevistas assim como da análise do material coletado nos fóruns, chats, canal \"Fale Conosco\" e demais espaços de registro e circulação de dados presentes em cada um dos sites. Uma das conclusões da pesquisa aponta para o fato de que a \"dialogicidade\" e a \"interatividade\" são atributos que dependem mais da natureza da relação que se estabelece entre os mediadores dos processos educativos e os cursistas do que propriamente dos instrumentos tecnológicos disponibilizados. / This dissertation investigates how internet plays its part at educommunication projects implemented by NCE Núcleo de Comunicação e Educação da Universidade de São Paulo, in order to understand how this institution uses virtual language at the different programs implemented since 2001, near education public nets. This research has been developed based on a sample composed by sites of the following projects: Educom.TV (2002), TodeOlho.TV (2002), Educom.rádio (2001-2004) and Educomradio.centro-oeste (2005-2006). This job aims to identify the specific use of digital language at present and virtual education. As methodology, the researcher used interviews as well as analisys of various materials collected from foruns, chats, Talk to us channels and other registers and data spaces near meantioned sites. One of the conclusions of this research points that dialogue and interactivity are atributes that depend on the nature of the relationship established between the mediators of educational processes and the students than the technological instruments available.

A presente dissertação procura investigar que papel a Internet exerce nos projetos educomunicativos implementados pelo NCE Núcleo de Comunicação e Educação da Universidade de São Paulo, buscando entender como o Núcleo vem se apropriando da linguagem virtual nos diferentes programas implementados a partir de 2001, junto a redes públicas de educação. Para tanto, a pesquisa toma como amostragem os sites dos projetos Educom.TV (2002), TodeOlho.TV (2002), Educom.rádio (2001-2004) e Educomradio.centro-oeste (2005-2006). O trabalho busca identificar a especificidade do emprego da linguagem digital tanto no âmbito da formação presencial quanto no da formação a distância. Como metodologia, a pesquisa faz uso de entrevistas assim como da análise do material coletado nos fóruns, chats, canal \"Fale Conosco\" e demais espaços de registro e circulação de dados presentes em cada um dos sites. Uma das conclusões da pesquisa aponta para o fato de que a \"dialogicidade\" e a \"interatividade\" são atributos que dependem mais da natureza da relação que se estabelece entre os mediadores dos processos educativos e os cursistas do que propriamente dos instrumentos tecnológicos disponibilizados. / This dissertation investigates how internet plays its part at educommunication projects implemented by NCE Núcleo de Comunicação e Educação da Universidade de São Paulo, in order to understand how this institution uses virtual language at the different programs implemented since 2001, near education public nets. This research has been developed based on a sample composed by sites of the following projects: Educom.TV (2002), TodeOlho.TV (2002), Educom.rádio (2001-2004) and Educomradio.centro-oeste (2005-2006). This job aims to identify the specific use of digital language at present and virtual education. As methodology, the researcher used interviews as well as analisys of various materials collected from foruns, chats, Talk to us channels and other registers and data spaces near meantioned sites. One of the conclusions of this research points that dialogue and interactivity are atributes that depend on the nature of the relationship established between the mediators of educational processes and the students than the technological instruments available.

This thesis formulates an approach towards the creation of a gesture activated and body movement controlled real time virtual 3d world in a dance performance context. It investigates immersion and navigation techniques derived from modern video games and methodologies and proposes how they can be used to further involve a performer into a virtual space as well as simultaneously offer a stimulating visual spectacle for an audience. The argument presented develops through practice-based methodology and artistic production strategies in interdisciplinary and collaborative contexts. Two choreographic performance/installations are used as cases studies to demonstrate in practice the proposed methodologies. First, the interactive dance work Suna No Onna, created in collaboration with Birringer/Danjoux and the Dap Lab, investigates the use of interactive pre-rendered animations in a real time setting and in real time by incorporating wearable sensors in the performance. Secondly, the potentials offered by the sensor technology and real time rendering engines led to the “creation scene", a key scene in the choreographic installation UKIYO (Moveable Worlds). This thesis investigates the design, creation and interaction qualities of virtual 3d spaces by exploring the potentialities offered by a shared space, between an intelligent space and a dancer in a hybrid world. The methodology applied uses as a theoretical base the phenomenological approach of Merleau-Ponty and Mark Hansen‟s mixed reality paradigm proposing the concept of the “space schema", a system which replicates and embeds proprioception, perception and motility into the space fabric offering a world which “lives”, functions and interacts with the performer. The outcome of the research is the generation of an interactive, non-linear, randomized 3d virtual space that collaborates with a technologically embedded performer in creating a 3d world which evolves and transforms, driven by the performer‟s intention and agency. This research contributes to the field of interactive performance art by making transparent the methodology, the instruments and the code used, in a non-technical terminology, making it accessible for both team members with less technological expertise as well as artists aspiring to engage interactive 3d media promoting further experimentation and conceptual discussions.

Submitted by biblioteca unifenas (biblioteca@unifenas.br) on 2017-08-24T13:10:32Z No. of bitstreams: 1 Carolina Soares Horta de Souza.pdf: 898395 bytes, checksum: 8025b51d63570bf92cd6485289fa3c65 (MD5) / Made available in DSpace on 2017-08-24T13:10:32Z (GMT). No. of bitstreams: 1 Carolina Soares Horta de Souza.pdf: 898395 bytes, checksum: 8025b51d63570bf92cd6485289fa3c65 (MD5) Previous issue date: 2015-02-25 / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - CAPES / The addition of the food coloring plays an important role in the food industry under the technological point of view. Natural dyes are difficult to navigate in standard products, so we were created synthetic dyes. Among artificial synthetic dyes allowed by law, are those of azo group and its use is one of the most controversial developments in the sector. The use of these substances is still raising a lot of questions as to toxicity, because in the literature, working with this momentum is limited and controversial. In this context, the aim, with this study was to evaluate the genotoxic potential of azo dyes by the micronucleus assay (MN). The MN test is widely used and internationally accepted for evaluation of mutagenic actions. The in vivo genotoxicity test was applied to research observed in erythrocytes Polychromatic (PCE) extracted from the bone marrow of the femur 168 Swiss mice Albinus, male and female, underwent five treatments (N-nitroso-N-ethylurea: control+; 150 mM NaCl: Control–; dye at 1, 1.5 and 2 g.Kg-1) and used four azo dyes (Tartrazine, Sunset Yellow, Red 40, Ponceau 4R) for 24h and 48h (after this time euthanasia was performed). We analyzed also the PCE/NCE ratio, which it is an important marker of cytotoxicity. The data obtained in the MN assay were subjected to analysis of one-way variance (ANOVA) using a factorial scheme of 5 × 2 × 2 (treatment × sex × time), and average compared to Tukey's test (p < 0.05) using SAS® software version 9.3. The dye Tartrazine showed genotoxicity dose- and sex-dependent; Sunset Yellow was genotoxic sex-independent; Red 40, its genotoxicity was dependent on the exposure time and the sex of the animal; the dye Ponceau 4R genotoxicity depends on the dose and time until euthanasia. However, it can be concluded that studies of these food additives often must be performed in order to constantly update data that are safe for consumption, since all four dyes tested in this study showed some degree of toxicity. Recalling that they have the use regulated by specific legislation establishing the Maximum Allowable Limits (LMP) and patterns of Acceptable Daily Intake for humans, extrapolated to animals. However, despite the control required by regulatory agencies, the use of dyes in food continues to raise a number of questions, mainly due to lack of studies evaluating the toxicity of these compounds which reinforces the attention they should be given. / A adição de corantes em alimentos e rações assume papel importante na indústria alimentícia sob o ponto de vista tecnológico. Corantes naturais são difíceis de serem utilizados em produtos padronizados, portanto, foram criados corantes sintéticos. Dentre os corantes sintéticos artificiais permitidos por lei, estão aqueles do grupo Azo e seu uso é um dos avanços mais controversos no setor. A utilização destas substâncias continua levantando uma série de dúvidas quanto à toxicidade, uma vez que, na literatura, trabalhos com esse ímpeto são escassos e controversos. Neste contexto, objetivou-se, com este trabalho, avaliar o potencial genotóxico de corantes azóicos através do ensaio de micronúcleo (MN). O teste do MN é amplamente utilizado e aceito internacionalmente para avaliação de ações mutagênicas. O teste in vivo foi aplicado para investigação de genotoxicidade observada em Eritrócitos Policromáticos (PCE) extraídos da medula óssea do fêmur de 168 camundongos Swiss albinus, machos e fêmeas, submetidos a cinco tratamentos (N-Nitroso-N-etilureia: controle+; 150 mM NaCl: controle–; corante a 1, 1,5 e 2 g.Kg-1), sendo utilizados quatro corantes azóicos (Tartrazina, Amarelo Crepúsculo, Vermelho 40, Ponceau 4R) nos tempos de 24h e 48h (após estes períodos foi executada eutanásia). Foi analisada também a relação PCE/NCE, cujo se trata de um importante biomarcador de citotoxicidade. Os dados obtidos no ensaio de MN foram submetidos à análise de variância (ANOVA), utilizando um esquema fatorial de 5 × 2 × 2 (tratamento × sexo × tempo), e comparação das médias pelo teste de Tukey (p < 0,05) usando o programa computacional SAS® versão 9.3. O corante Tartrazina apresentou genotoxicidade dose e sexo dependentes; Amarelo Crepúsculo foi genotóxico independente do sexo; Vermelho 40, a sua genotoxicidade foi dependente do tempo de exposição e do sexo do animal; a genotoxicidade do corante Ponceau 4R foi dependente da dose administrada e do tempo até a eutanásia. Contudo, pode-se concluir que estudos envolvendo estes aditivos alimentares devem ser realizados frequentemente com a finalidade de atualização constante de dados que ofereçam segurança para seu consumo, uma vez que todos os quatro corantes testados nesta pesquisa apresentaram algum grau de toxicidade. Lembrando que os mesmos possuem o uso regulamentado por legislação específica, que estabelece os Limites Máximos Permitidos (LMP) e padrões de Ingestão Diária Aceitável por humanos, extrapolados para animais. Porém, apesar do controle exigido pelas agências reguladoras, a utilização de corantes em alimentos continua levantando uma série de dúvidas, principalmente devido a escassez de trabalhos avaliando a toxicidade destes compostos o que reforça a atenção que lhes deve ser conferida.

Unnormalised parametric models are an important class of probabilistic models which are difficult to estimate. The models are important since they occur in many different areas of application, e.g. in modelling of natural images, natural language and associative memory. However, standard maximum likelihood estimation is not applicable to unnormalised models, so alternative methods are required. Noise-contrastive estimation (NCE) has been proposed as an effective estimation method for unnormalised models. The basic idea is to transform the unsupervised estimation problem into a supervised classification problem. The parameters of the unnormalised model are learned by training the model to differentiate the given data samples from generated noise samples. However, the choice of the noise distribution has been left open to the user, and as the performance of the estimation may be sensitive to this choice, it is desirable for it to be automated. In this thesis, the ambiguity in the choice of the noise distribution is addressed by presenting the previously unpublished conditional noise-contrastive estimation (CNCE) method. Like NCE, CNCE estimates unnormalised models by classifying data and noise samples. However, the choice of noise distribution is partly automated via the use of a conditional noise distribution that is dependent on the data. In addition to introducing the core theory for CNCE, the method is empirically validated on data and models where the ground truth is known. Furthermore, CNCE is applied to natural image data to show its applicability in a realistic application. / Icke-normaliserade parametriska modeller utgör en viktig klass av svåruppskattade statistiska modeller. Dessa modeller är viktiga eftersom de uppträder inom många olika tillämpningsområden, t.ex. vid modellering av bilder, tal och skrift och associativt minne. Dessa modeller är svåruppskattade eftersom den vanliga maximum likelihood-metoden inte är tillämpbar på icke-normaliserade modeller. Noise-contrastive estimation (NCE) har föreslagits som en effektiv metod för uppskattning av icke-normaliserade modeller. Grundidén är att transformera det icke-handledda uppskattningsproblemet till ett handlett klassificeringsproblem. Den icke-normaliserade modellens parametrar blir inlärda genom att träna modellen på att skilja det givna dataprovet från ett genererat brusprov. Dock har valet av brusdistribution lämnats öppet för användaren. Eftersom uppskattningens prestanda är känslig gentemot det här valet är det önskvärt att få det automatiserat. I det här examensarbetet behandlas valet av brusdistribution genom att presentera den tidigare opublicerade metoden conditional noise-contrastive estimation (CNCE). Liksom NCE uppskattar CNCE icke-normaliserade modeller via klassificering av data- och brusprov. I det här fallet är emellertid brusdistributionen delvis automatiserad genom att använda en betingad brusdistribution som är beroende på dataprovet. Förutom att introducera kärnteorin för CNCE valideras även metoden med hjälp av data och modeller vars genererande parametrar är kända. Vidare appliceras CNCE på bilddata för att demonstrera dess tillämpbarhet.