Supervised Aggregative Feature Extraction for Big Data Time Series Regression

In many applications, and especially those where batch processes are involved, a target scalar output of interest is often dependent on one or more time series of data. With the exponential growth in data logging in modern industries such time series are increasingly available for statistical modeling in soft sensing applications. In order to exploit time series data for predictive modelling, it is necessary to summarise the information they contain as a set of features to use as model regressors. Typically this is done in an unsupervised fashion using simple techniques such as computing statistical moments, principal components or wavelet decompositions, often leading to significant information loss and hence suboptimal predictive models. In this paper, a functional learning paradigm is exploited in a supervised fashion to derive continuous, smooth estimates of time series data (yielding aggregated local information), while simultaneously estimating a continuous shape function yielding optimal predictions. The proposed Supervised Aggregative Feature Extraction (SAFE) methodology can be extended to support nonlinear predictive models by embedding the functional learning framework in a Reproducing Kernel Hilbert Spaces setting. SAFE has a number of attractive features including closed form solution and the ability to explicitly incorporate first and second order derivative information. Using simulation studies and a practical semiconductor manufacturing case study we highlight the strengths of the new methodology with respect to standard unsupervised feature extraction approaches.

A Fast Algorithm for Sparse Support Vector Machines for Mobile Computing Applications

This research presents a fast algorithm for projected support vector machines (PSVM) by selecting a basis vector set (BVS) for the kernel-induced feature space, the training points are projected onto the subspace spanned by the selected BVS. A standard linear support vector machine (SVM) is then produced in the subspace with the projected training points. As the dimension of the subspace is determined by the size of the selected basis vector set, the size of the produced SVM expansion can be specified. A two-stage algorithm is derived which selects and refines the basis vector set achieving a locally optimal model. The model expansion coefficients and bias are updated recursively for increase and decrease in the basis set and support vector set. The condition for a point to be classed as outside the current basis vector and selected as a new basis vector is derived and embedded in the recursive procedure. This guarantees the linear independence of the produced basis set. The proposed algorithm is tested and compared with an existing sparse primal SVM (SpSVM) and a standard SVM (LibSVM) on seven public benchmark classification problems. Our new algorithm is designed for use in the application area of human activity recognition using smart devices and embedded sensors where their sometimes limited memory and processing resources must be exploited to the full and the more robust and accurate the classification the more satisfied the user. Experimental results demonstrate the effectiveness and efficiency of the proposed algorithm. This work builds upon a previously published algorithm specifically created for activity recognition within mobile applications for the EU Haptimap project [1]. The algorithms detailed in this paper are more memory and resource efficient making them suitable for use with bigger data sets and more easily trained SVMs.

Acceleration-as-a-Service: Exploiting Virtualised GPUs for a Financial Application

How can GPU acceleration be obtained as a service in a cluster? This question has become increasingly significant due to the inefficiency of installing GPUs on all nodes of a cluster. The research reported in this paper is motivated to address the above question by employing rCUDA (remote CUDA), a framework that facilitates Acceleration-as-a-Service (AaaS), such that the nodes of a cluster can request the acceleration of a set of remote GPUs on demand. The rCUDA framework exploits virtualisation and ensures that multiple nodes can share the same GPU. In this paper we test the feasibility of the rCUDA framework on a real-world application employed in the financial risk industry that can benefit from AaaS in the production setting. The results confirm the feasibility of rCUDA and highlight that rCUDA achieves similar performance compared to CUDA, provides consistent results, and more importantly, allows for a single application to benefit from all the GPUs available in the cluster without loosing efficiency.

Wide-Area Phase-Angle Measurements for Islanding Detection—An Adaptive Nonlinear Approach

The integration of an ever growing proportion of large scale distributed renewable generation has increased the probability of maloperation of the traditional RoCoF and vector shift relays. With reduced inertia due to non-synchronous penetration in a power grid, system wide disturbances have forced the utility industry to design advanced protection schemes to prevent system degradation and avoid cascading outages leading to widespread blackouts. This paper explores a novel adaptive nonlinear approach applied to islanding detection, based on wide area phase angle measurements. This is challenging, since the voltage phase angles from different locations exhibit not only strong nonlinear but also time-varying characteristics. The adaptive nonlinear technique, called moving window kernel principal component analysis is proposed to model the time-varying and nonlinear trends in the voltage phase angle data. The effectiveness of the technique is exemplified using both DigSilent simulated cases and real test cases recorded from the Great Britain and Ireland power systems by the OpenPMU project.

Modelling the fresh properties of self-compacting concrete using support vector machine approach

The main objective of the study presented in this paper was to investigate the feasibility using support vector machines (SVM) for the prediction of the fresh properties of self-compacting concrete. The radial basis function (RBF) and polynomial kernels were used to predict these properties as a function of the content of mix components. The fresh properties were assessed with the slump flow, T50, T60, V-funnel time, Orimet time, and blocking ratio (L-box). The retention of these tests was also measured at 30 and 60 min after adding the first water. The water dosage varied from 188 to 208 L/m3, the dosage of superplasticiser (SP) from 3.8 to 5.8 kg/m3, and the volume of coarse aggregates from 220 to 360 L/m3. In total, twenty mixes were used to measure the fresh state properties with different mixture compositions. RBF kernel was more accurate compared to polynomial kernel based support vector machines with a root mean square error (RMSE) of 26.9 (correlation coefficient of R2 = 0.974) for slump flow prediction, a RMSE of 0.55 (R2 = 0.910) for T50 (s) prediction, a RMSE of 1.71 (R2 = 0.812) for T60 (s) prediction, a RMSE of 0.1517 (R2 = 0.990) for V-funnel time prediction, a RMSE of 3.99 (R2 = 0.976) for Orimet time prediction, and a RMSE of 0.042 (R2 = 0.988) for L-box ratio prediction, respectively. A sensitivity analysis was performed to evaluate the effects of the dosage of cement and limestone powder, the water content, the volumes of coarse aggregate and sand, the dosage of SP and the testing time on the predicted test responses. The analysis indicates that the proposed SVM RBF model can gain a high precision, which provides an alternative method for predicting the fresh properties of SCC.

On the Privacy of Encrypted Skype Communications

The privacy of voice over IP (VoIP) systems is achieved by compressing and encrypting the sampled data. This paper investigates in detail the leakage of information from Skype, a widely used VoIP application. In this research, it has been demonstrated by using the dynamic time warping (DTW) algorithm, that sentences can be identified with an accuracy of 60%. The results can be further improved by choosing specific training data. An approach involving the Kalman filter is proposed to extract the kernel of all training signals.

Hα Line Profile Asymmetries and the Chromospheric Flare Velocity Field

The asymmetries observed in the line profiles of solar flares can provide important diagnostics of the properties and dynamics of the flaring atmosphere. In this paper the evolution of the Hα and Ca ii λ8542 lines are studied using high spatial, temporal, and spectral resolution ground-based observations of an M1.1 flare obtained with the Swedish 1 m Solar Telescope. The temporal evolution of the Hα line profiles from the flare kernel shows excess emission in the red wing (red asymmetry) before flare maximum and excess in the blue wing (blue asymmetry) after maximum. However, the Ca ii λ8542 line does not follow the same pattern, showing only a weak red asymmetry during the flare. RADYN simulations are used to synthesize spectral line profiles for the flaring atmosphere, and good agreement is found with the observations. We show that the red asymmetry observed in Hα is not necessarily associated with plasma downflows, and the blue asymmetry may not be related to plasma upflows. Indeed, we conclude that the steep velocity gradients in the flaring chromosphere modify the wavelength of the central reversal in the Hα line profile. The shift in the wavelength of maximum opacity to shorter and longer wavelengths generates the red and blue asymmetries, respectively.

Face Recognition in the Scrambled Domain via Salience-Aware Ensembles of Many Kernels

With the rapid development of internet-of-things (IoT), face scrambling has been proposed for privacy protection during IoT-targeted image/video distribution. Consequently in these IoT applications, biometric verification needs to be carried out in the scrambled domain, presenting significant challenges in face recognition. Since face models become chaotic signals after scrambling/encryption, a typical solution is to utilize traditional data-driven face recognition algorithms. While chaotic pattern recognition is still a challenging task, in this paper we propose a new ensemble approach – Many-Kernel Random Discriminant Analysis (MK-RDA) to discover discriminative patterns from chaotic signals. We also incorporate a salience-aware strategy into the proposed ensemble method to handle chaotic facial patterns in the scrambled domain, where random selections of features are made on semantic components via salience modelling. In our experiments, the proposed MK-RDA was tested rigorously on three human face datasets: the ORL face dataset, the PIE face dataset and the PUBFIG wild face dataset. The experimental results successfully demonstrate that the proposed scheme can effectively handle chaotic signals and significantly improve the recognition accuracy, making our method a promising candidate for secure biometric verification in emerging IoT applications.

Subspace-based techniques and applications

Este trabalho focou-se no estudo de técnicas de sub-espaço tendo em vista as aplicações seguintes: eliminação de ruído em séries temporais e extracção de características para problemas de classificação supervisionada. Foram estudadas as vertentes lineares e não-lineares das referidas técnicas tendo como ponto de partida os algoritmos SSA e KPCA. No trabalho apresentam-se propostas para optimizar os algoritmos, bem como uma descrição dos mesmos numa abordagem diferente daquela que é feita na literatura. Em qualquer das vertentes, linear ou não-linear, os métodos são apresentados utilizando uma formulação algébrica consistente. O modelo de subespaço é obtido calculando a decomposição em valores e vectores próprios das matrizes de kernel ou de correlação/covariância calculadas com um conjunto de dados multidimensional. A complexidade das técnicas não lineares de subespaço é discutida, nomeadamente, o problema da pre-imagem e a decomposição em valores e vectores próprios de matrizes de dimensão elevada. Diferentes algoritmos de préimagem são apresentados bem como propostas alternativas para a sua optimização. A decomposição em vectores próprios da matriz de kernel baseada em aproximações low-rank da matriz conduz a um algoritmo mais eficiente- o Greedy KPCA. Os algoritmos são aplicados a sinais artificiais de modo a estudar a influência dos vários parâmetros na sua performance. Para além disso, a exploração destas técnicas é extendida à eliminação de artefactos em séries temporais biomédicas univariáveis, nomeadamente, sinais EEG.

A study of singular integral operators with shift

Nesta tese, consideram-se operadores integrais singulares com a acção extra de um operador de deslocacamento de Carleman e com coeficientes em diferentes classes de funções essencialmente limitadas. Nomeadamente, funções contínuas por troços, funções quase-periódicas e funções possuíndo factorização generalizada. Nos casos dos operadores integrais singulares com deslocamento dado pelo operador de reflexão ou pelo operador de salto no círculo unitário complexo, obtêm-se critérios para a propriedade de Fredholm. Para os coeficientes contínuos, uma fórmula do índice de Fredholm é apresentada. Estes resultados são consequência das relações de equivalência explícitas entre aqueles operadores e alguns operadores adicionais, tais como o operador integral singular, operadores de Toeplitz e operadores de Toeplitz mais Hankel. Além disso, as relações de equivalência permitem-nos obter um critério de invertibilidade e fórmulas para os inversos laterais dos operadores iniciais com coeficientes factorizáveis. Adicionalmente, aplicamos técnicas de análise numérica, tais como métodos de colocação de polinómios, para o estudo da dimensão do núcleo dos dois tipos de operadores integrais singulares com coeficientes contínuos por troços. Esta abordagem permite também a computação do inverso no sentido Moore-Penrose dos operadores principais. Para operadores integrais singulares com operadores de deslocamento do tipo Carleman preservando a orientação e com funções contínuas como coeficientes, são obtidos limites superiores da dimensão do núcleo. Tal é implementado utilizando algumas estimativas e com a ajuda de relações (explícitas) de equivalência entre operadores. Focamos ainda a nossa atenção na resolução e nas soluções de uma classe de equações integrais singulares com deslocamento que não pode ser reduzida a um problema de valor de fronteira binomial. De forma a atingir os objectivos propostos, foram utilizadas projecções complementares e identidades entre operadores. Desta forma, as equações em estudo são associadas a sistemas de equações integrais singulares. Estes sistemas são depois analisados utilizando um problema de valor de fronteira de Riemann. Este procedimento tem como consequência a construção das soluções das equações iniciais a partir das soluções de problemas de valor de fronteira de Riemann. Motivados por uma grande diversidade de aplicações, estendemos a definição de operador integral de Cauchy para espaços de Lebesgue sobre grupos topológicos. Assim, são investigadas as condições de invertibilidade dos operadores integrais neste contexto.

Regularity of Wiener-Hopf plus Hankel operators

In this thesis we consider Wiener-Hopf-Hankel operators with Fourier symbols in the class of almost periodic, semi-almost periodic and piecewise almost periodic functions. In the first place, we consider Wiener-Hopf-Hankel operators acting between L2 Lebesgue spaces with possibly different Fourier matrix symbols in the Wiener-Hopf and in the Hankel operators. In the second place, we consider these operators with equal Fourier symbols and acting between weighted Lebesgue spaces Lp(R;w), where 1 < p < 1 and w belongs to a subclass of Muckenhoupt weights. In addition, singular integral operators with Carleman shift and almost periodic coefficients are also object of study. The main purpose of this thesis is to obtain regularity properties characterizations of those classes of operators. By regularity properties we mean those that depend on the kernel and cokernel of the operator. The main techniques used are the equivalence relations between operators and the factorization theory. An invertibility characterization for the Wiener-Hopf-Hankel operators with symbols belonging to the Wiener subclass of almost periodic functions APW is obtained, assuming that a particular matrix function admits a numerical range bounded away from zero and based on the values of a certain mean motion. For Wiener-Hopf-Hankel operators acting between L2-spaces and with possibly different AP symbols, criteria for the semi-Fredholm property and for one-sided and both-sided invertibility are obtained and the inverses for all possible cases are exhibited. For such results, a new type of AP factorization is introduced. Singular integral operators with Carleman shift and scalar almost periodic coefficients are also studied. Considering an auxiliar and simpler operator, and using appropriate factorizations, the dimensions of the kernels and cokernels of those operators are obtained. For Wiener-Hopf-Hankel operators with (possibly different) SAP and PAP matrix symbols and acting between L2-spaces, criteria for the Fredholm property are presented as well as the sum of the Fredholm indices of the Wiener-Hopf plus Hankel and Wiener-Hopf minus Hankel operators. By studying dependencies between different matrix Fourier symbols of Wiener-Hopf plus Hankel operators acting between L2-spaces, results about the kernel and cokernel of those operators are derived. For Wiener-Hopf-Hankel operators acting between weighted Lebesgue spaces, Lp(R;w), a study is made considering equal scalar Fourier symbols in the Wiener-Hopf and in the Hankel operators and belonging to the classes of APp;w, SAPp;w and PAPp;w. It is obtained an invertibility characterization for Wiener-Hopf plus Hankel operators with APp;w symbols. In the cases for which the Fourier symbols of the operators belong to SAPp;w and PAPp;w, it is obtained semi-Fredholm criteria for Wiener-Hopf-Hankel operators as well as formulas for the Fredholm indices of those operators.

Reflections of universal algebras into semilattices, their Galois theories, and related factorization systems

Estabelecemos uma condição suficiente para a preservação dos produtos finitos, pelo reflector de uma variedade de álgebras universais numa subvariedade, que é, também, condição necessária se a subvariedade for idempotente. Esta condição é estabelecida, seguidamente, num contexto mais geral e caracteriza reflexões para as quais a propriedade de ser semi-exacta à esquerda e a propriedade, mais forte, de ter unidades estáveis, coincidem. Prova-se que reflexões simples e semi-exactas à esquerda coincidem, no contexto das variedades de álgebras universais e caracterizam-se as classes do sistema de factorização derivado da reflexão. Estabelecem-se resultados que ajudam a caracterizar morfismos de cobertura e verticais-estáveis em álgebras universais e no contexto mais geral já referido. Caracterizam-se as classes de morfismos separáveis, puramente inseparáveis e normais. O estudo dos morfismos de descida de Galois conduz a condições suficientes para que o seu par kernel seja preservado pelo reflector.

A time frequency approach to blind deconvolution in multipath underwater channels

Blind deconvolution is studied in the underwater acoustic channel context, by time-frequency (TF) processing. The acoustic propagation environment is modelled by ray tracing and mathematically described by a multipath propagation channel. Representation of the received signal by means of a signal-dependent TF distribution (radially Gaussian kernel distribution) allowed to visualize the resolved replicas of the emitted signal, while signi cantly attenuating the inherent interferences of classic quadratic TF distributions. The source signal instantaneous frequency estimation was the starting point for both source and channel estimation. Source signature estimation was performed by either TF inversion, based on the Wigner-Ville distribution of the received signal, or a subspace- -based method. The channel estimate was obtained either via a TF formulation of the conventional matched- lter, or via matched- - ltering with the previously obtained source estimate. A shallow water realistic scenario is considered, comprising a 135-m depth water column and an acoustic source located at 90-m depth and 5.6-km range from the receiver. For the corresponding noiseless simulated data, the quality of the best estimates was 0.856 for the source signal, and 0.9664 and 0.9996 for the amplitudes and time-delays of the impulse response, respectively. Application of the proposed deconvolution method to real data of the INTIMATE '96 sea trial conduced to source and channel estimates with the quality of 0.530 and 0.843, respectively. TF processing has proved to remove the typical ill-conditioning of single sensor deterministic deconvolution techniques.

Automatic Video Analysis for Fisheries Survey Systems

Thesis (Ph.D.)--University of Washington, 2015

Modelo de data mining para deteção de embolias pulmonares

Trabalho de Projeto para obtenção do grau de Mestre em Engenharia Informática e de Computadores