A contribução original desta tese é um novo algoritmo que integra um aproximador de funções com alta eficiência amostral com aprendizagem por reforço em espaços de estados contínuos. A pesquisa completa inclui o desenvolvimento de um algoritmo online e incremental capaz de aprender por meio de uma única passada sobre os dados. Este algoritmo, chamado de Fast Incremental Gaussian Mixture Network (FIGMN) foi empregado como um aproximador de funções eficiente para o espaço de estados de tarefas contínuas de aprendizagem por reforço, que, combinado com Q-learning linear, resulta em performance competitiva. Então, este mesmo aproximador de funções foi empregado para modelar o espaço conjunto de estados e valores Q, todos em uma única FIGMN, resultando em um algoritmo conciso e com alta eficiência amostral, i.e., um algoritmo de aprendizagem por reforço capaz de aprender por meio de pouquíssimas interações com o ambiente. Um único episódio é suficiente para aprender as tarefas investigadas na maioria dos experimentos. Os resultados são analisados a fim de explicar as propriedades do algoritmo obtido, e é observado que o uso da FIGMN como aproximador de funções oferece algumas importantes vantagens para aprendizagem por reforço em relação a redes neurais convencionais. / This thesis’ original contribution is a novel algorithm which integrates a data-efficient function approximator with reinforcement learning in continuous state spaces. The complete research includes the development of a scalable online and incremental algorithm capable of learning from a single pass through data. This algorithm, called Fast Incremental Gaussian Mixture Network (FIGMN), was employed as a sample-efficient function approximator for the state space of continuous reinforcement learning tasks, which, combined with linear Q-learning, results in competitive performance. Then, this same function approximator was employed to model the joint state and Q-values space, all in a single FIGMN, resulting in a concise and data-efficient algorithm, i.e., a reinforcement learning algorithm that learns from very few interactions with the environment. A single episode is enough to learn the investigated tasks in most trials. Results are analysed in order to explain the properties of the obtained algorithm, and it is observed that the use of the FIGMN function approximator brings some important advantages to reinforcement learning in relation to conventional neural networks.

Classificação de fluxos contínuos de dados possui muitos desafios para a comunidade de mineração de dados quando o ambiente não é estacionário. Um dos maiores desafios para a aprendizagem em fluxos contínuos de dados está relacionado com a adaptação às mudanças de conceito, as quais ocorrem como resultado da evolução dos dados ao longo do tempo. Duas formas principais de desenvolver abordagens adaptativas são os métodos baseados em conjunto de classificadores e os algoritmos incrementais. Métodos baseados em conjunto de classificadores desempenham um papel importante devido à sua modularidade, o que proporciona uma maneira natural de se adaptar a mudanças de conceito. Os algoritmos incrementais são mais rápidos e possuem uma melhor capacidade anti-ruído do que os conjuntos de classificadores, mas têm mais restrições sobre os fluxos de dados. Assim, é um desafio combinar a flexibilidade e a adaptação de um conjunto de classificadores na presença de mudança de conceito, com a simplicidade de uso encontrada em um único classificador com aprendizado incremental. Com essa motivação, nesta dissertação, propomos um algoritmo incremental, online e probabilístico para a classificação em problemas que envolvem mudança de conceito. O algoritmo é chamado IGMN-NSE e é uma adaptação do algoritmo IGMN. As duas principais contribuições da IGMN-NSE em relação à IGMN são: melhoria de poder preditivo para tarefas de classificação e a adaptação para alcançar um bom desempenho em cenários não estacionários. Estudos extensivos em bases de dados sintéticas e do mundo real demonstram que o algoritmo proposto pode rastrear os ambientes em mudança de forma muito próxima, independentemente do tipo de mudança de conceito. / Data stream classification poses many challenges for the data mining community when the environment is non-stationary. The greatest challenge in learning classifiers from data stream relates to adaptation to the concept drifts, which occur as a result of changes in the underlying concepts. Two main ways to develop adaptive approaches are ensemble methods and incremental algorithms. Ensemble method plays an important role due to its modularity, which provides a natural way of adapting to change. Incremental algorithms are faster and have better anti-noise capacity than ensemble algorithms, but have more restrictions on concept drifting data streams. Thus, it is a challenge to combine the flexibility and adaptation of an ensemble classifier in the presence of concept drift, with the simplicity of use found in a single classifier with incremental learning. With this motivation, in this dissertation we propose an incremental, online and probabilistic algorithm for classification as an effort of tackling concept drifting. The algorithm is called IGMN-NSE and is an adaptation of the IGMN algorithm. The two main contributions of IGMN-NSE in relation to the IGMN are: predictive power improvement for classification tasks and adaptation to achieve a good performance in non-stationary environments. Extensive studies on both synthetic and real-world data demonstrate that the proposed algorithm can track the changing environments very closely, regardless of the type of concept drift.

This thesis proposes a probabilistic level set method to be used in segmentation of tumors with heterogeneous intensities. It models the intensities of the tumor and surrounding tissue using Gaussian mixture models. Through a contour based initialization procedure samples are gathered to be used in expectation maximization of the mixture model parameters. The proposed method is compared against a threshold-based segmentation method using MRI images retrieved from The Cancer Imaging Archive. The cases are manually segmented and an automated testing procedure is used to find optimal parameters for the proposed method and then it is tested against the threshold-based method. Segmentation times, dice coefficients, and volume errors are compared. The evaluation reveals that the proposed method has a comparable mean segmentation time to the threshold-based method, and performs faster in cases where the volume error does not exceed 40%. The mean dice coefficient and volume error are also improved while achieving lower deviation.

Using Gaussian mixtures is a popular and very flexible approach to statistical modelling. The standard approach of maximum likelihood estimation cannot be used for some of these models. The estimates are, however, obtainable by iterative solutions, such as the EM (Expectation-Maximization) algorithm. The aim of this thesis is to present Gaussian mixture models and their implementation in R. The non-trivial case of having to use the EM algorithm is assumed. Existing methods and packages are presented, investigated and compared. Some of them are extended by custom R code. Several exhaustive simulations are run and some of the interesting results are presented. For these simulations, a notion of usual fit is presented.

South Africa is currently struggling to deal with a significant poaching and livestock theft problem. This work is concerned with the detection and classification of ground based targets using radar micro- Doppler signatures to aid in the monitoring of borders, nature reserves and farmlands. The research starts of by investigating the state of the art of ground target classification. Different radar systems are investigated with respect to their ability to classify targets at different operating frequencies. Finally, a Gaussian Mixture Model Hidden Markov Model based (GMM-HMM) classification approach is presented and tested in an operational environment. The GMM-HMM method is compared to methods in the literature and is shown to achieve reasonable (up to 95%) classification accuracy, marginally outperforming existing ground target classification methods. / Dissertation (MEng)--University of Pretoria, 2017. / Electrical, Electronic and Computer Engineering / MEng / Unrestricted

<p>U ovoj disertaciji je predstavljen model koji aproksimira pune kova-<br />rijansne matrice u modelu gausovih mešavina (GMM) sa smanjenim<br />brojem parametara i izračunavanja koji su potrebni za izračunavanje<br />izglednosti. U predloženom modelu inverzne kovarijansne matrice su<br />aproksimirane korišćenjem retke reprezentacije njihovih karakteri-<br />stičnih vektora. Pored samog modela prikazan je i algoritam za<br />estimaciju parametara zasnovan na kriterijumu maksimizacije<br />izgeldnosti. Eksperimentalni rezultati na problemu prepoznavanja<br />govora su pokazali da predloženi model za isti nivo greške kao GMM<br />sa upunim kovarijansnim, redukuje broj parametara za 45%.</p> / <p>This thesis proposes a model which approximates full covariance matrices in<br />Gaussian mixture models with a reduced number of parameters and<br />computations required for likelihood evaluations. In the proposed model<br />inverse covariance (precision) matrices are approximated using sparsely<br />represented eigenvectors. A maximum likelihood algorithm for parameter<br />estimation and its practical implementation are presented. Experimental<br />results on a speech recognition task show that while keeping the word error<br />rate close to the one obtained by GMMs with full covariance matrices, the<br />proposed model can reduce the number of parameters by 45%.</p>

With the shift toward online shopping, it has become necessary to customize customers' needs and give them more choices. Before making a purchase, buyers research the products' features. The recommender systems facilitate the search task for customers by narrowing down the search space within specific products that align with the customer's needs. A recommender system uses clustering to filter information, calculating the similarity between members of a cluster to determine the factors that will lead to more accurate predictions. We propose a new method for predicting scores in machine learning recommender systems using the Gaussian mixture model clustering and the Pearson correlation coefficient. The proposed method is applied to MovieLens data. The results are then compared to three commonly used methods: Pearson correlation coefficients, K-means, and fuzzy C-means algorithms. As a result of increasing the number of neighbors, our method shows a lower error than others. Additionally, the results depict that accuracy will increase as the number of users increases. Our model, for instance, is 5% more accurate than existing methods when the neighbor size is 30. Gaussian mixture clustering chooses similar users and takes into account the scores distance when choosing nearby users that are similar to the active user.

Trucks are designed, configured and marketed for various working environments. There lies a concern whether trucks are used as intended by the manufacturer, as usage may impact the longevity, efficiency and productivity of the trucks. In this thesis we propose a framework divided into two separate parts, that aims to extract costumers’ driving behaviours from Logged Vehicle Data (LVD) in order to a): evaluate whether they align with so-called Global Transport Application (GTA) parameters and b): evaluate the usage in terms of performance. Gaussian mixture model (GMM) is employed to cluster and classify various driving behaviors. Association rule mining was applied on the categorized clusters to validate that the usage follow GTA configuration. Furthermore, Correlation Coefficient (CC) was used to find linear relationships between usage and performance in terms of Fuel Consumption (FC). It is found that the vast majority of the trucks seemingly follow GTA parameters, thus used as marketed. Likewise, the fuel economy was found to be linearly dependent with drivers’ various performances. The LVD lacks detail, such as Global Positioning System (GPS) information, needed to capture the usage in such a way that more definitive conclusions can be drawn. / <p>This thesis was later conducted as a scientific paper and was submit- ted to the conference of ICIMP, 2020. The publication was accepted the 23th of September (2019), and will be presented in January, 2020.</p>

Cette thèse traite de ce verrou théorique pour des vidéos naturelles. Les contributions scientifiques développées ont permis : 1. De réfuter mathématiquement le modèle gaussien en général adopté dans la littérature pour représenter le bruit de canal ; 2. D’établir pour la première fois, le caractère stationnaire des processus aléatoires représentant le bruit de canal, la méthode développée étant indépendante du type de données, de leur traitement et de la procédure d’estimation ; 3. De proposer une méthodologie de modélisation du bruit de canal à partir d’un mélange de gaussiennes pour une transformée aussi bien en cosinus discrète qu’en ondelette discrète et pour un large ensemble d’attaques (filtrage, rotation, compression, StirMark, …). L’intérêt de cette approche est entre autres de permettre le calcul exact de la capacité du canal alors que la littérature ne fournissait que des bornes supérieure et inférieure. 4. Les contributions technologique concernent l’intégration et l’implémentions de ces modèles dans la méthode du tatouage IProtect brevetée Institut Télécom/ARTEMIS et SFR avec un gain en temps d’exécution d’un facteur 100 par rapport à l’état de l’art. / The thesis is focus on natural video and attack modelling for uncompressed video watermarking purposes. By reconsidering a statistical investigation combining four types of statistical tests, the thesis starts by identifying with accuracy the drawbacks and limitations of the popular Gaussian model in watermarking applications. Further on, an advanced statistical approach is developed in order to establish with mathematical rigour: 1. That a mathematical model for the original video content and/or attacks exists; 2. The model parameters. From the theoretical point of view, this means to prove for the first time the stationarity of the random processes representing the natural video and/or the watermarking attacks. These general results have been already validated under applicative and theoretical frameworks. On the one hand, when integrating the attack models into the IProtect watermarking method patented by Institut Télécom/ARTEMIS and SFR, a speed-up by a factor of 100 of the insertion procedure has been obtained. On the other hand, accurate models for natural video and attacks allowed the increasing of the precision in the computation of some basic information theory entities (entropies and capacity).

Missing data is a common issue in data analysis. Numerous techniques have been proposed to deal with the missing data problem. Imputation is the most popular strategy for handling the missing data. Imputation for data analysis is the process to replace the missing values with any plausible values. Two most frequent imputation techniques cited in literature are the single imputation and the multiple imputation. The multiple imputation, also known as the golden imputation technique, has been proposed by Rubin in 1987 to address the missing data. However, the inconsistency is the major problem in the multiple imputation technique. The single imputation is less popular in missing data research due to bias and less variability issues. One of the solutions to improve the single imputation technique in the basic regression model: the main motivation is that, the residual is added to improve the bias and variability. The residual is drawn by normal distribution assumption with a mean of 0, and the variance is equal to the residual variance. Although new methods in the single imputation technique, such as stochastic regression model, and hot deck imputation, might be able to improve the variability and bias issues, the single imputation techniques suffer with the uncertainty that may underestimate the R-square or standard error in the analysis results. The research reported in this thesis provides two imputation solutions for the single imputation technique. In the first imputation procedure, the wild bootstrap is proposed to improve the uncertainty for the residual variance in the regression model. In the second solution, the predictive mean matching (PMM) is enhanced, where the regression model is taking the main role to generate the recipient values while the observations in the donors are taken from the observed values. Then the missing values are imputed by randomly drawing one of the observations in the donor pool. The size of the donor pool is significant to determine the quality of the imputed values. The fixed size of donor is used to be employed in many existing research works with PMM imputation technique, but might not be appropriate in certain circumstance such as when the data distribution has high density region. Instead of using the fixed size of donor pool, the proposed method applies the radius-based solution to determine the size of donor pool. Both proposed imputation procedures will be combined with the Gaussian mixture model framework to preserve the original data distribution. The results reported in the thesis from the experiments on benchmark and artificial data sets confirm improvement for further data analysis. The proposed approaches are therefore worthwhile to be considered for further investigation and experiments.