MESQUITA, Diego Parente Paiva. Machine Learning for incomplete data. 2017. 55 f. Dissertação (Mestrado em Ciência da Computação)-Universidade Federal do Ceará, Fortaleza, 2017. / Submitted by Jonatas Martins (jonatasmartins@lia.ufc.br) on 2017-08-29T14:42:43Z No. of bitstreams: 1 2017_dis_dppmesquita.pdf: 673221 bytes, checksum: eec550f75e2965d1120185327465a595 (MD5) / Approved for entry into archive by Rocilda Sales (rocilda@ufc.br) on 2017-08-29T16:04:36Z (GMT) No. of bitstreams: 1 2017_dis_dppmesquita.pdf: 673221 bytes, checksum: eec550f75e2965d1120185327465a595 (MD5) / Made available in DSpace on 2017-08-29T16:04:36Z (GMT). No. of bitstreams: 1 2017_dis_dppmesquita.pdf: 673221 bytes, checksum: eec550f75e2965d1120185327465a595 (MD5) Previous issue date: 2017 / Methods based on basis functions (such as the sigmoid and q-Gaussian functions) and similarity measures (such as distances or kernel functions) are widely used in machine learning and related fields. These methods often take for granted that data is fully observed and are not equipped to handle incomplete data in an organic manner. This assumption is often flawed, as incomplete data is a fact in various domains such as medical diagnosis and sensor analytics. Therefore, one might find it useful to be able to estimate the value of these functions in the presence of partially observed data. We propose methodologies to estimate the Gaussian Kernel, the Euclidean Distance, the Epanechnikov kernel and arbitrary basis functions in the presence of possibly incomplete feature vectors. To obtain such estimates, the incomplete feature vectors are treated as continuous random variables and, based on that, we take the expected value of the transforms of interest. / Métodos baseados em funções de base (como as funções sigmoid e a q-Gaussian) e medidas de similaridade (como distâncias ou funções de kernel) são comuns em Aprendizado de Máquina e áreas correlatas. Comumente, no entanto, esses métodos não são equipados para utilizar dados incompletos de maneira orgânica. Isso pode ser visto como um impedimento, uma vez que dados parcialmente observados são comuns em vários domínios, como aplicações médicas e dados provenientes de sensores. Nesta dissertação, propomos metodologias para estimar o valor do kernel Gaussiano, da distância Euclidiana, do kernel Epanechnikov e de funções de base arbitrárias na presença de vetores possivelmente parcialmente observados. Para obter tais estimativas, os vetores incompletos são tratados como variáveis aleatórias contínuas e, baseado nisso, tomamos o valor esperado da transformada de interesse.

Automatic speech recognition (ASR) is a well-researched field of study aimed at augmenting the man-machine interface through interpretation of the spoken word. From in-car voice recognition systems to automated telephone directories, automatic speech recognition technology is becoming increasingly abundant in today’s technological world. Nonetheless, traditional audio-only ASR system performance degrades when employed in noisy environments such as moving vehicles. To improve system performance under these conditions, visual speech information can be incorporated into the ASR system, yielding what is known as audio-video speech recognition (AVASR). A majority of AVASR research focuses on lip parameters extraction within controlled environments, but these scenarios fail to meet the demanding requirements of most real-world applications. Within the visual unconstrained environment, AVASR systems must compete with constantly changing lighting conditions and background clutter as well as subject movement in three dimensions. This work proposes a robust still image lip localization algorithm capable of operating in an unconstrained visual environment, serving as a visual front end to AVASR systems. A novel Bhattacharyya-based face detection algorithm is used to compare candidate regions of interest with a unique illumination-dependent face model probability distribution function approximation. Following face detection, a lip-specific Gabor filter-based feature space is utilized to extract facial features and localize lips within the frame. Results indicate a 75% lip localization overall success rate despite the demands of the visually noisy environment.

Submitted by José Ignacio Valencia Díaz (jivalenciadiaz@gmail.com) on 2013-09-17T00:13:33Z No. of bitstreams: 1 Dissertacao MPFE Jose Ignacio Valencia Diaz.pdf: 1741345 bytes, checksum: b45af943bf4f6e8a2a9963c07038d9dc (MD5) / Approved for entry into archive by Suzinei Teles Garcia Garcia (suzinei.garcia@fgv.br) on 2013-09-17T12:05:59Z (GMT) No. of bitstreams: 1 Dissertacao MPFE Jose Ignacio Valencia Diaz.pdf: 1741345 bytes, checksum: b45af943bf4f6e8a2a9963c07038d9dc (MD5) / Made available in DSpace on 2013-09-17T12:54:35Z (GMT). No. of bitstreams: 1 Dissertacao MPFE Jose Ignacio Valencia Diaz.pdf: 1741345 bytes, checksum: b45af943bf4f6e8a2a9963c07038d9dc (MD5) Previous issue date: 2013-08-26 / Prediction models based on nonparametric estimation are in continuous development and have been permeating the quantitative community. Their main feature is that they do not consider as known a priori the form of the probability distributions functions (PDF), but allow the data to be used directly in order to build their own PDFs. In this work it is implemented the nonparametric pooled estimators from Sam and Jiang (2009) for drift and diffusion functions for the short rate diffusion process, by means of the use of yield series of different maturities provided by One Day Future Interbank Deposit contracts (ID1). The estimators are built from the perspective of kernel functions and they are optimized with a particular kernel format, in our case, Epanechnikov’s kernel, and with a smoothing parameter (bandwidth). Empiric experience indicates that the smoothing parameter is critical to find the probability density function that provides an optimal estimation in terms of MISE (Mean Integrated Squared Error) when testing the model with the traditional k-folds cross-validation method. Exceptions arise when the series do not have appropriate sizes, but the structural break of the diffusion process of the Brazilian interest short rate, since 2006, requires the reduction of the length of the series to the cost of reducing the predictive power of the model. This structural break represents the evolution of the Brazilian market, in an attempt to converge towards mature markets and it explains largely the unsatisfactory performance of the proposed estimator. / Modelos de predição baseados em estimações não-paramétricas continuam em desenvolvimento e têm permeado a comunidade quantitativa. Sua principal característica é que não consideram a priori distribuições de probabilidade conhecidas, mas permitem que os dados passados sirvam de base para a construção das próprias distribuições. Implementamos para o mercado brasileiro os estimadores agrupados não-paramétricos de Sam e Jiang (2009) para as funções de drift e de difusão do processo estocástico da taxa de juros instantânea, por meio do uso de séries de taxas de juros de diferentes maturidades fornecidas pelos contratos futuros de depósitos interfinanceiros de um dia (DI1). Os estimadores foram construídos sob a perspectiva da estimação por núcleos (kernels), que requer para a sua otimização um formato específico da função-núcleo. Neste trabalho, foi usado o núcleo de Epanechnikov, e um parâmetro de suavizamento (largura de banda), o qual é fundamental para encontrar a função de densidade de probabilidade ótima que forneça a estimação mais eficiente em termos do MISE (Mean Integrated Squared Error - Erro Quadrado Integrado Médio) no momento de testar o modelo com o tradicional método de validação cruzada de k-dobras. Ressalvas são feitas quando as séries não possuem os tamanhos adequados, mas a quebra estrutural do processo de difusão da taxa de juros brasileira, a partir do ano 2006, obriga à redução do tamanho das séries ao custo de reduzir o poder preditivo do modelo. A quebra estrutural representa um processo de amadurecimento do mercado brasileiro que provoca em grande medida o desempenho insatisfatório do estimador proposto.