No description available.

The dissertation deals with clustering algorithms and transforming regression prob-lems into classification problems. The main contributions of the dissertation are twofold; first, to improve (speed up) the clustering algorithms and second, to develop a strict learn-ing environment for solving regression problems as classification tasks by using support vector machines (SVMs). An extension to the most popular unsupervised clustering meth-od, k-means algorithm, is proposed, dubbed k-means2 (k-means squared) algorithm, appli-cable to ultra large datasets. The main idea is based on using a small portion of the dataset in the first stage of the clustering. Thus, the centers of such a smaller dataset are computed much faster than if computing the centers based on the whole dataset. These final centers of the first stage are naturally much closer to the locations of the final centers rendering a great reduction in the total computational cost. For large datasets the speed up in computa-tion exhibited a trend which is shown to be high and rising with the increase in the size of the dataset. The total transient time for the fast stage was found to depend largely on the portion of the dataset selected in the stage. For medium size datasets it has been shown that an 8-10% portion of data used in the fast stage is a reasonable choice. The centers of the 8-10% samples computed during the fast stage may oscillate towards the final centers' positions of the fast stage along the centers' movement path. The slow stage will start with the final centers of the fast phase and the paths of the centers in the second stage will be much shorter than the ones of a classic k-means algorithm. Additionally, the oscillations of the slow stage centers' trajectories along the path to the final centers' positions are also greatly minimized. In the second part of the dissertation, a novel approach of posing a solution of re-gression problems as the multiclass classification tasks within the common framework of kernel machines is proposed. Based on such an approach both the nonlinear (NL) regression problems and NL multiclass classification tasks will be solved as multiclass classification problems by using SVMs. The accuracy of an approximating classification (hyper)Surface (averaged over several benchmarking data sets used in this study) to the data points over a given high-dimensional input space created by a nonlinear multiclass classifier is slightly superior to the solution obtained by regression (hyper)Surface. In terms of the CPU time needed for training (i.e. for tuning the hyperparameters of the models), the nonlinear SVM classifier also shows significant advantages. Here, the comparisons between the solutions obtained by an SVM solving given regression problem as a classic SVM regressor and as the SVM classifier have been performed. In order to transform a regression problem into a classification task, four possible discretizations of a continuous output (target) vector y are introduced and compared. A very strict double (nested) cross-validation technique has been used for measuring the performances of regression and multiclass classification SVMs. In order to carry out fair comparisons, SVMs are used for solving both tasks - regression and multiclass classification. The readily available and most popular benchmarking SVM tool, LibSVM, was used in all experiments. The results in solving twelve benchmarking regression tasks shown here will present SVM regression and classification algorithms as strongly competing models where each approach shows merits for a specific class of high-dimensional function approximation problems.

Made available in DSpace on 2014-06-11T19:22:34Z (GMT). No. of bitstreams: 0 Previous issue date: 2011-12-13Bitstream added on 2014-06-13T20:09:49Z : No. of bitstreams: 1 geronimo_tm_me_bauru.pdf: 1704386 bytes, checksum: 1a6dd612ef17da8f95f721e29761eddc (MD5) / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES) / O monitoramento de processos de fabricação pro usinagem tem se mostrado de extrema importância nas empresas que buscam um nível de excelência em qualidade, auxiliando na melhor alocação de recursos e redução de desperdícios oriundos de peças com problemas de qualidade. Abordagens multisensoriais têm sido empregadas no monitoramento desses processos com o objetivo de utilizar os sinais no treinamento de sistemas de inteligência artificial na tarefa de indicar desvios nas ferramentas ou no produto sendo fabricado. Neste trabalho, três sistemas de inteligência artificial foram utilizados com o o objetivo de prover estimativas para o diâmetro de furos obtidos por processo de furação de precisão. Uma rede neural artificial perceptron de múltiplas camadas (RNA MLP), um sistema de inerferência adaptável neuro-fuzzy (ANFIS) e uma rede neural artificial com função de base radial (RBF) foram treinados com os dados obtidos com os sensores para estimar os diâmetros dos furos para cada material de corpo-de-prova. A definição da melhor configuração para cada sistema de inteligência artificial foi obtida através de algoritmos desenvolvidos para verificar a influência dos sinais e dos parâmetros particulares de cada sistema sobre o resultado final da estimativa. Os resultados obtidos indicam que a RNA MLP apresenta maior robutez perante a variação nos dados apresentados. O sistema ANFIS e a rede RFB mostraram que seu resultado varia acentuadamente quando há variações nos dados apresentados no treinamento, requerendo que estes sistemas sejam treinados sempre com o conjunto de dados apresentados na mesma ordem. A análise de influência dos sinais mostrou que, embora a abordagem multisensorial apresente bons resultados na rede MLP, o mesmo não se repetiu para os demais sistemas... / The supervision of manufacturing process by machining has been extremely important in companies which aim an excellence level in quality, helping on best assets allocation and waste reduction originated from pieces with quality problems. Multi-sensory approaches have been employed in the supervision of these processes aiming the use of signals in the training of artificial intelligence systems in order to indicate deviations in tools or in product being manufactured. Turning, grinding, milling and drilling have benn the target of the application of these supervision intelligence systems. In this work these artificial intelligence systems were applied in order to provide estimations for the diameters of the holes obtained by precision drilling process. A Multilayer Perceptron Neural Network (ANN MLP), and adaptive neuro-fuzzy inference system (ANFIS) and an artificial neural network with radial basis function (RBF) were trained with the data obtained from the sensors to estimate the hole diameters for each material of the test pieces. The definition of the best configuration for each artificial intelligence system was obtained through algorithms developed to verify the signals influence and particular parameters of each system concerning the final estimation result. The tests results were analyzed under three criteria: the absolute and medium errors, the system capacity of obtaining correct results for each estimation - classifying them as approved or rejected - and the error analysis regarding the necessary tolerance classes to maintain process within the limits of precision mechanics. The results obtained indicate that the ANN MLP presents higher robustness before variation in the data presented. The ANFIS system and RFB network have shown that their result vary sharply when there are data variations presented in training... (Complete abstract click electronic access below)

В данной работе на тему «Разработка симулятора виртуальной реальности», произведена разработка и сборка прототипа автосимулятора грузового автомобиля, за основу программы для персонального компьютера использовалось приложение Euro track Simulator 2. В результате проделанной работы, получился симулятор, не уступающий по своему функционалу аналогам на рынке, но с более низкой стоимостью. А также произведено внедрение устройства в автошколу. / In this work on the topic "Development of a virtual reality simulator", a prototype of a truck simulator was developed and assembled, the Euro track Simulator 2 application was taken as the basis for the program. As a result of the work done, we received a simulator that is not inferior in functionality to analogues on the market, but at a lower cost. And also the device was implemented in a driving school.

Made available in DSpace on 2016-04-26T18:18:07Z (GMT). No. of bitstreams: 1 Eduardo Kiochi Nakamiti.pdf: 508081 bytes, checksum: 823ace52fadc0ffa1e477b48ec078c0f (MD5) Previous issue date: 2009-09-01 / The diffusion of Artificial Intelligent Agents, superficially considered as elements that have independent thinking and whose decisions influence processes in the everyday human reality increasingly known, especially in electronic media, the users of banking services, on access to telephone service in our home appliances and entertainment. Its presence is sometimes felt by the ease of access to resources often so powerful and sometimes so unnoticed. It can be perceived in a negative way, when access to electronic banking services is denied without explanation. Much of learning today, relies on Internet and getting information online without the help of search engines is disappointing. Today, these mechanisms offer more than simply the result of a search. They seem to have the intelligence to offer suggestions in relation to our interest. This form of communication that are incorporating to their customs, is strongly influenced by known agents, in computer systems. In this work, we will make a history of its appearance, and technological innovations and cultural, primarily focusing on the first Artificial Intelligence in order to built a clear, detailed view of the Artificial Intelligent Agents, at present, and to establish a hypothetical path for future developments. The methodological basis of work is based on the complexity of the issue addressed. Such complexity refers to tools and distinct views, dynamic and interactive, pointing to the computer and its various aspects, the artificial intelligence and its tools, the cyber arena of interdisciplinarity as the first and, to the environment and the semiotic analysis tool of the production and mediation of knowledge and theories of complexity. The corpus of the analysis and interpretation are circumscribed to the phenomenon of Artificial Intelligent Agents of nowadays and of the future, but for both, and according to the methodology adopted, the fields of knowledge are inspected more in the expectation of higher fidelity description. The main conclusion is the remarkable trend of spraying and increasing invisibility of Artificial Intelligent Agents, as elements of support in decision making and control of the environment and the information we received / A explosão dos Agentes Inteligentes Artificiais, considerados superficialmente como elementos que apresentam raciocínio autônomo e cujas decisões influem processos, é uma realidade cada vez mais sentida no cotidiano dos seres humanos , principalmente nos meios de comunicação eletrônicos, entre usuários de serviços bancários, em acessos a serviços telefônicos, nos eletrodomésticos e em nossos entretenimentos. Sua presença é sentida às vezes pela facilidade de acesso a recursos muitas vezes de forma poderosa ou outras vezes de forma discreta. Também pode ser sentida de forma negativa, quando ocorre o bloqueio de uma operação bancária eletrônica sem explicação. Muito do aprendizado, hoje em dia, passa pela internet, e buscar informação na rede sem a ajuda de mecanismos de busca é desconcertante. Hoje em dia esses mecanismos nos oferecem mais do que simplesmente o resultado de uma procura. Eles parecem ter inteligência própria ao nos oferecer sugestões relacionadas ao nosso interesse. Esta forma de comunicação está se incorporando aos costumes e é fortemente influenciada pelos denominados agentes, contidos nos sistemas. Nesse trabalho, é feito um histórico do seu aparecimento e das inovações tecnológicas e culturais introduzidas, focalizando, primariamente, os primórdios da Inteligência Artificial para que seja construída uma visão clara e detalhada dos agentes inteligentes artificiais até o momento presente, a fim de que seja possível estabelecer uma trajetória hipotética de evolução futura. O referencial teórico do trabalho apóia-se na complexidade do tema abordado. Complexidade que evoca ferramentas e olhares distintos, dinâmicos e interatuantes, apontando para a computação e seus vários aspectos, a inteligência artificial e suas ferramentas; a cibernética como primeira arena de interdisciplinaridade e, atingindo a semiótica como ambiente e ferramenta de análise do processo de produção e mediação do conhecimento e as teorias da complexidade. O corpus da análise e interpretação restringe-se ao fenômeno dos Agentes Inteligentes Artificiais desde sua origem, a partir da inspeção de vários campos de conhecimento, necessários à compreensão do tema abordado. O trabalho levanta a hipótese da tendência marcante de pulverização e invisibilidade crescentes dos Agentes Inteligentes Artificiais, como elementos de apoio na tomada de decisão e controle da informação recebida

Orientador: Prof. Dr. Fabrício Olivetti de França / Dissertação (mestrado) - Universidade Federal do ABC, Programa de Pós-Graduação em Ciência da Computação, 2016. / No campo de Inteligência Artficial uma dasareas de interesse é a de criação de agentes inteligentes. Essa area de estudo tem o objetivo de construir um agente capaz de tomar decisões sem intervenção humana, para cumprir determinadas tarefas. Um desafio dessa área é manter o bom cumprimento da tarefa quando o agente enfrenta situações distintas do que ele observou durante a fase do aprendizado. Uma das possíveis aplicações desses agentesé na área de Jogos Eletrronicos, em que agentes autonomos podem ser programados para vencer os desafios projetados pelos programadores, com o objetivo de verificar a viabilidade e dificuldade de cada estagio. Da mesma forma, os jogos eletrronicos podem ser utilizados como plataforma para criação de novos algoritmos de aprendizado para agentes autonomos a serem aplicados em situações diversas. Este trabalho proproe um ambiente de teste chamado EvoMan, onde os oito inimigo finais do jogo eletronico MegaMan II 1 foram reproduzidos e estão disponíveis para treinar agentes controladores articiais de diferentes formas, simulando a dinâmica de um ambiente incerto. Os experimentos compreendem o uso de algoritmos de NeuroEvolução para encontrar conjuntos de agentes autônomos capazes de enfrentar os diversos desaos interpostos no ambiente proposto de forma generalista.

Orientador: Paulo Roberto de Aguiar / Banca: Eraldo Janinone da Silva / Banca: José Alfredo Covolan Ulson / Resumo: O monitoramento de processos de fabricação pro usinagem tem se mostrado de extrema importância nas empresas que buscam um nível de excelência em qualidade, auxiliando na melhor alocação de recursos e redução de desperdícios oriundos de peças com problemas de qualidade. Abordagens multisensoriais têm sido empregadas no monitoramento desses processos com o objetivo de utilizar os sinais no treinamento de sistemas de inteligência artificial na tarefa de indicar desvios nas ferramentas ou no produto sendo fabricado. Neste trabalho, três sistemas de inteligência artificial foram utilizados com o o objetivo de prover estimativas para o diâmetro de furos obtidos por processo de furação de precisão. Uma rede neural artificial perceptron de múltiplas camadas (RNA MLP), um sistema de inerferência adaptável neuro-fuzzy (ANFIS) e uma rede neural artificial com função de base radial (RBF) foram treinados com os dados obtidos com os sensores para estimar os diâmetros dos furos para cada material de corpo-de-prova. A definição da melhor configuração para cada sistema de inteligência artificial foi obtida através de algoritmos desenvolvidos para verificar a influência dos sinais e dos parâmetros particulares de cada sistema sobre o resultado final da estimativa. Os resultados obtidos indicam que a RNA MLP apresenta maior robutez perante a variação nos dados apresentados. O sistema ANFIS e a rede RFB mostraram que seu resultado varia acentuadamente quando há variações nos dados apresentados no treinamento, requerendo que estes sistemas sejam treinados sempre com o conjunto de dados apresentados na mesma ordem. A análise de influência dos sinais mostrou que, embora a abordagem multisensorial apresente bons resultados na rede MLP, o mesmo não se repetiu para os demais sistemas... (Resumo completo, clicar acesso eletrônico abaixo) / Abstract: The supervision of manufacturing process by machining has been extremely important in companies which aim an excellence level in quality, helping on best assets allocation and waste reduction originated from pieces with quality problems. Multi-sensory approaches have been employed in the supervision of these processes aiming the use of signals in the training of artificial intelligence systems in order to indicate deviations in tools or in product being manufactured. Turning, grinding, milling and drilling have benn the target of the application of these supervision intelligence systems. In this work these artificial intelligence systems were applied in order to provide estimations for the diameters of the holes obtained by precision drilling process. A Multilayer Perceptron Neural Network (ANN MLP), and adaptive neuro-fuzzy inference system (ANFIS) and an artificial neural network with radial basis function (RBF) were trained with the data obtained from the sensors to estimate the hole diameters for each material of the test pieces. The definition of the best configuration for each artificial intelligence system was obtained through algorithms developed to verify the signals influence and particular parameters of each system concerning the final estimation result. The tests results were analyzed under three criteria: the absolute and medium errors, the system capacity of obtaining correct results for each estimation - classifying them as approved or rejected - and the error analysis regarding the necessary tolerance classes to maintain process within the limits of precision mechanics. The results obtained indicate that the ANN MLP presents higher robustness before variation in the data presented. The ANFIS system and RFB network have shown that their result vary sharply when there are data variations presented in training... (Complete abstract click electronic access below) / Mestre

В рамках работы были исследованы методы построения систем планирования и управления предприятием класса ERP-систем. Цель работы – разработка программного инструмента в «1С: ERP Управление предприятием 2», позволяющего вести учет всех заказов, предстоящих и уже выполняющихся на предприятии. В ходе исследования был разработан рабочий прототип учета и планирования производства заказов на предприятии по производству литий-ионных аккумуляторов с использованием ERP-системы «1С-Предприятие». / Within the framework of this work, a study was made of the existing types of scheduler and enterprise resources planning system methods. An analysis of the existing question-answer systems was carried out. The goal of work id development software tool for schedule and resources planning for order control and management of laity-ions buttery base on ERP-system “1C: ERP Enterprise”. In the course of the study, a working prototype of schedule and resources planning for order control and management base on ERP-system “1C: ERP Enterprise” was developed.

В рамках данной работы было проведено исследование существующих типов вопросно-ответных систем и методов анализа текста. Был проведен анализ существующих вопросно-ответных систем. Приведена обобщённая схема работы вопросно-ответных систем и для каждого типа систем приведена детальная схема работы. Описаны и исследованы методы выбора кандидатов ответа и методы их оценки. Также в работе описаны возможные критерии оценки работы таких систем. В ходе исследования был разработан рабочий прототип вопросно-ответной системы, основанный на системе BERT для русского языка. Используемая модель RuBERT была предобучена и протестирована на стандартных задачах SQuAD. В ходе работы модель была протестирована и оценена в разработанном рабочем прототипе и показала высокие результаты по предложенным критериям оценки. / Within the framework of this work, a study was made of the existing types of question-answer systems and text analysis methods. An analysis of the existing question-answer systems was carried out. A generalized scheme of operation of question-answer systems is given, and a detailed scheme of operation is given for each type of system. Methods for selecting response candidates and methods for their evaluation are described and investigated. The paper also describes possible criteria for evaluating the operation of such systems. In the course of the study, a working prototype of a question-answer system based on the BERT system for the Russian language was developed. The RuBERT model used was pre-trained and tested on standard SQuAD problems. During the work, the model was tested and evaluated in the developed working prototype and showed high results according to the proposed evaluation criteria.

No description available.