Development of Directionally Adaptive Techniques for Single Image Super-Resolution

Super Resolution problem is an inverse problem and refers to the process of producing a High resolution (HR) image, making use of one or more Low Resolution (LR) observations. It includes up sampling the image, thereby, increasing the maximum spatial frequency and removing degradations that arise during the image capture namely aliasing and blurring. The work presented in this thesis is based on learning based single image super-resolution. In learning based super-resolution algorithms, a training set or database of available HR images are used to construct the HR image of an image captured using a LR camera. In the training set, images are stored as patches or coefficients of feature representations like wavelet transform, DCT, etc. Single frame image super-resolution can be used in applications where database of HR images are available. The advantage of this method is that by skilfully creating a database of suitable training images, one can improve the quality of the super-resolved image. A new super resolution method based on wavelet transform is developed and it is better than conventional wavelet transform based methods and standard interpolation methods. Super-resolution techniques based on skewed anisotropic transform called directionlet transform are developed to convert a low resolution image which is of small size into a high resolution image of large size. Super-resolution algorithm not only increases the size, but also reduces the degradations occurred during the process of capturing image. This method outperforms the standard interpolation methods and the wavelet methods, both visually and in terms of SNR values. Artifacts like aliasing and ringing effects are also eliminated in this method. The super-resolution methods are implemented using, both critically sampled and over sampled directionlets. The conventional directionlet transform is computationally complex. Hence lifting scheme is used for implementation of directionlets. The new single image super-resolution method based on lifting scheme reduces computational complexity and thereby reduces computation time. The quality of the super resolved image depends on the type of wavelet basis used. A study is conducted to find the effect of different wavelets on the single image super-resolution method. Finally this new method implemented on grey images is extended to colour images and noisy images

Using Analogy to Acquire Commonsense Knowledge from Human Contributors

The goal of the work reported here is to capture the commonsense knowledge of non-expert human contributors. Achieving this goal will enable more intelligent human-computer interfaces and pave the way for computers to reason about our world. In the domain of natural language processing, it will provide the world knowledge much needed for semantic processing of natural language. To acquire knowledge from contributors not trained in knowledge engineering, I take the following four steps: (i) develop a knowledge representation (KR) model for simple assertions in natural language, (ii) introduce cumulative analogy, a class of nearest-neighbor based analogical reasoning algorithms over this representation, (iii) argue that cumulative analogy is well suited for knowledge acquisition (KA) based on a theoretical analysis of effectiveness of KA with this approach, and (iv) test the KR model and the effectiveness of the cumulative analogy algorithms empirically. To investigate effectiveness of cumulative analogy for KA empirically, Learner, an open source system for KA by cumulative analogy has been implemented, deployed, and evaluated. (The site "1001 Questions," is available at http://teach-computers.org/learner.html). Learner acquires assertion-level knowledge by constructing shallow semantic analogies between a KA topic and its nearest neighbors and posing these analogies as natural language questions to human contributors. Suppose, for example, that based on the knowledge about "newspapers" already present in the knowledge base, Learner judges "newspaper" to be similar to "book" and "magazine." Further suppose that assertions "books contain information" and "magazines contain information" are also already in the knowledge base. Then Learner will use cumulative analogy from the similar topics to ask humans whether "newspapers contain information." Because similarity between topics is computed based on what is already known about them, Learner exhibits bootstrapping behavior --- the quality of its questions improves as it gathers more knowledge. By summing evidence for and against posing any given question, Learner also exhibits noise tolerance, limiting the effect of incorrect similarities. The KA power of shallow semantic analogy from nearest neighbors is one of the main findings of this thesis. I perform an analysis of commonsense knowledge collected by another research effort that did not rely on analogical reasoning and demonstrate that indeed there is sufficient amount of correlation in the knowledge base to motivate using cumulative analogy from nearest neighbors as a KA method. Empirically, evaluating the percentages of questions answered affirmatively, negatively and judged to be nonsensical in the cumulative analogy case compares favorably with the baseline, no-similarity case that relies on random objects rather than nearest neighbors. Of the questions generated by cumulative analogy, contributors answered 45% affirmatively, 28% negatively and marked 13% as nonsensical; in the control, no-similarity case 8% of questions were answered affirmatively, 60% negatively and 26% were marked as nonsensical.

Aplicació d'algoritmes genètics en l'optimització dels processos de fabricació del paper

La creciente preocupación y concienciación de la sociedad respecto el medio ambiente, y en consecuencia la legislación y regulaciones generadas inducen a la modificación de los procesos productivos existentes en la industria química. Las configuraciones iniciales deben modificarse para conseguir una mayor integración de procesos. Para este fin se han creado y desarrollado diferentes metodologías que deben facilitar la tarea a los responsables del rediseño. El desarrollo de una metodología y herramientas complementarias es el principal objetivo de la investigación aquí presentada, especialmente centrada en el desarrollo y la aplicación de una metodología de optimización de procesos. Esta metodología de optimización se aplica sobre configuraciones de proceso existentes y pretende encontrar nuevas configuraciones viables según los objetivos de optimización fijados. La metodología tiene dos partes diferenciadas: la primera se basa en un simulador de procesos comercial y la segunda es la técnica de optimización propiamente dicha. La metodología se inicia con la elaboración de una simulación convenientemente validada que reproduzca el proceso existente, en este caso una papelera no integrada que produce papel estucado de calidad, para impresión. A continuación la técnica de optimización realiza una búsqueda dentro del dominio de los posibles resultados, en busca de los mejores resultados que satisfazcan plenamente los objetivos planteados. Dicha técnica de optimización está basada en los algoritmos genéticos como herramienta de búsqueda, junto a un subprograma basado en técnicas de programación matemática para el cálculo de resultados. Un número reducido de resultados son finalmente escogidos y utilizados para modificar la simulación existente fijando la redistribución de los flujos del proceso. Los resultados de la simulación del proceso determinan en último caso la viabilidad técnica de cada reconfiguración planteada. En el proceso de optimización, los objetivos están definidos en una función objetivo dentro de la técnica de optimización. Dicha función rige la búsqueda de resultados. La función objetivo puede ser individual o una combinación de objetivos. En el presente caso, la función persigue una minimización del consumo de agua y una minimización de la pérdida de materia prima. La optimización se realiza bajo restricciones para alcanzar este objetivo combinado en forma de una solución de compromiso. Producto de la aplicación de esta metodología se han obtenido resultados interesantes que significan una mejora del cierre de circuitos y un ahorro de materia prima, sin comprometer al mismo tiempo la operabilidad del proceso producto ni la calidad del papel.

Model selection approaches for non-linear system identification: a review

The identification of non-linear systems using only observed finite datasets has become a mature research area over the last two decades. A class of linear-in-the-parameter models with universal approximation capabilities have been intensively studied and widely used due to the availability of many linear-learning algorithms and their inherent convergence conditions. This article presents a systematic overview of basic research on model selection approaches for linear-in-the-parameter models. One of the fundamental problems in non-linear system identification is to find the minimal model with the best model generalisation performance from observational data only. The important concepts in achieving good model generalisation used in various non-linear system-identification algorithms are first reviewed, including Bayesian parameter regularisation and models selective criteria based on the cross validation and experimental design. A significant advance in machine learning has been the development of the support vector machine as a means for identifying kernel models based on the structural risk minimisation principle. The developments on the convex optimisation-based model construction algorithms including the support vector regression algorithms are outlined. Input selection algorithms and on-line system identification algorithms are also included in this review. Finally, some industrial applications of non-linear models are discussed.

A concept of approximated densities for efficient nonlinear estimation

This paper presents the theoretical development of a nonlinear adaptive filter based on a concept of filtering by approximated densities (FAD). The most common procedures for nonlinear estimation apply the extended Kalman filter. As opposed to conventional techniques, the proposed recursive algorithm does not require any linearisation. The prediction uses a maximum entropy principle subject to constraints. Thus, the densities created are of an exponential type and depend on a finite number of parameters. The filtering yields recursive equations involving these parameters. The update applies the Bayes theorem. Through simulation on a generic exponential model, the proposed nonlinear filter is implemented and the results prove to be superior to that of the extended Kalman filter and a class of nonlinear filters based on partitioning algorithms.

Motion estimation through approximated densities

Many techniques are currently used for motion estimation. In the block-based approaches the most common procedure applied is the block-matching based on various algorithms. To refine the motion estimates resulting from the full search or any coarse search algorithm, one can find few applications of Kalman filtering, mainly in the intraframe scheme. The Kalman filtering technique applicability for block-based motion estimation is rather limited due to discontinuities in the dynamic behaviour of the motion vectors. Therefore, we propose an application of the concept of the filtering by approximated densities (FAD). The FAD, originally introduced to alleviate limitations due to conventional Kalman modelling, is applied to interframe block-motion estimation. This application uses a simple form of FAD involving statistical characteristics of multi-modal distributions up to second order.

On merging gradient estimation with mean-tracking techniques for cluster identification

This paper discusses how numerical gradient estimation methods may be used in order to reduce the computational demands on a class of multidimensional clustering algorithms. The study is motivated by the recognition that several current point-density based cluster identification algorithms could benefit from a reduction of computational demand if approximate a-priori estimates of the cluster centres present in a given data set could be supplied as starting conditions for these algorithms. In this particular presentation, the algorithm shown to benefit from the technique is the Mean-Tracking (M-T) cluster algorithm, but the results obtained from the gradient estimation approach may also be applied to other clustering algorithms and their related disciplines.

New opportunities for secure communication networks using shaped femtosecond laser pulses inducing filamentation processes in the atmosphere

The current study discusses new opportunities for secure ground to satellite communications using shaped femtosecond pulses that induce spatial hole burning in the atmosphere for efficient communications with data encoded within super-continua generated by femtosecond pulses. Refractive index variation across the different layers in the atmosphere may be modelled using assumptions that the upper strata of the atmosphere and troposphere behaving as layered composite amorphous dielectric networks composed of resistors and capacitors with different time constants across each layer. Input-output expressions of the dynamics of the networks in the frequency domain provide the transmission characteristics of the propagation medium. Femtosecond pulse shaping may be used to optimize the pulse phase-front and spectral composition across the different layers in the atmosphere. A generic procedure based on evolutionary algorithms to perform the pulse shaping is proposed. In contrast to alternative procedures that would require ab initio modelling and calculations of the propagation constant for the pulse through the atmosphere, the proposed approach is adaptive, compensating for refractive index variations along the column of air between the transmitter and receiver.

Joint multiple dictionary learning for tensor sparse coding

Traditional dictionary learning algorithms are used for finding a sparse representation on high dimensional data by transforming samples into a one-dimensional (1D) vector. This 1D model loses the inherent spatial structure property of data. An alternative solution is to employ Tensor Decomposition for dictionary learning on their original structural form —a tensor— by learning multiple dictionaries along each mode and the corresponding sparse representation in respect to the Kronecker product of these dictionaries. To learn tensor dictionaries along each mode, all the existing methods update each dictionary iteratively in an alternating manner. Because atoms from each mode dictionary jointly make contributions to the sparsity of tensor, existing works ignore atoms correlations between different mode dictionaries by treating each mode dictionary independently. In this paper, we propose a joint multiple dictionary learning method for tensor sparse coding, which explores atom correlations for sparse representation and updates multiple atoms from each mode dictionary simultaneously. In this algorithm, the Frequent-Pattern Tree (FP-tree) mining algorithm is employed to exploit frequent atom patterns in the sparse representation. Inspired by the idea of K-SVD, we develop a new dictionary update method that jointly updates elements in each pattern. Experimental results demonstrate our method outperforms other tensor based dictionary learning algorithms.

Analysis of speech fluency in Williams syndrome

Williams syndrome (WS) is a neurodevelopmental genetic disorder, often referred as being characterized by dissociation between verbal and non-verbal abilities, although the number of studies disputing this proposal is emerging. Indeed, although they have been traditionally reported as displaying increased speech fluency, this topic has not been fully addressed in research. In previous studies carried out with a small group of individuals with WS, we reported speech breakdowns during conversational and autobiographical narratives suggestive of language difficulties. In the current study, we characterized the speech fluency profile using an ecologically based measure - a narrative task (story generation) was collected from a group of individuals with WS (n = 30) and typically developing group (n = 39) matched in mental age. Oral narratives were elicited using a picture stimulus - the cookie theft picture from Boston Diagnosis Aphasia Test. All narratives were analyzed according to typology and frequency of fluency breakdowns (non-stuttered and stuttered disfluencies). Oral narratives in WS group differed from typically developing group, mainly due to a significant increase in the frequency of disfluencies, particularly in terms of hesitations, repetitions and pauses. This is the first evidence of disfluencies in WS using an ecologically based task (oral narrative task), suggesting that these speech disfluencies may represent a significant marker of language problems in WS. (C) 2011 Elsevier Ltd. All rights reserved.

Particle Competition and Cooperation in Networks for Semi-Supervised Learning

Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP)

Alocação de canais em sistemas de comunicação celular empregando algoritmo genético distribuído

Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES)

Docking e análise do modo de ligação de três moléculas pequenas, um benzimidazol e dois compostos de crômio, nos sulcos do DNA 5'-CGCGAATTCGCG-3

Pós-graduação em Ciência e Tecnologia de Materiais - FC

Otimização multiobjetivo na análise da integração de geração distribuída às redes de distribuição

Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES)

Comparação de métodos baseados em algoritmos genéticos para ajuste coordenado de estabilizadores de sistemas de potência

Essa dissertação de mestrado apresenta um estudo comparativo entre três metodologias baseadas em algoritmos genéticos para ajuste coordenado de estabilizadores de sistemas de potência (ESP). Os procedimentos de ajuste do ESP são formulados como um problema de otimização, a fim de: 1) maximizar o coeficiente de amortecimento mínimo do sistema em malha fechada; 2) maximizar o somatório de todos os coeficientes de amortecimento do sistema em malha fechada; e 3) deslocar os modos eletromecânicos poucos amortecidos ou mal amortecidos para uma zona pré-escrita no plano s. As três metodologias consideram um conjunto de condições de operacionais pré-especificadas. O sistema elétrico foi representado por equações no espaço de estado e as matrizes associadas com a modelagem foram obtidas por meio da versão acadêmica do programa PacDyn. As simulações foram realizadas usando o MATLAB. As metodologias foram aplicadas no conhecido sistema teste New England.