Sparse Correlation Kernel Analysis and Reconstruction

This paper presents a new paradigm for signal reconstruction and superresolution, Correlation Kernel Analysis (CKA), that is based on the selection of a sparse set of bases from a large dictionary of class- specific basis functions. The basis functions that we use are the correlation functions of the class of signals we are analyzing. To choose the appropriate features from this large dictionary, we use Support Vector Machine (SVM) regression and compare this to traditional Principal Component Analysis (PCA) for the tasks of signal reconstruction, superresolution, and compression. The testbed we use in this paper is a set of images of pedestrians. This paper also presents results of experiments in which we use a dictionary of multiscale basis functions and then use Basis Pursuit De-Noising to obtain a sparse, multiscale approximation of a signal. The results are analyzed and we conclude that 1) when used with a sparse representation technique, the correlation function is an effective kernel for image reconstruction and superresolution, 2) for image compression, PCA and SVM have different tradeoffs, depending on the particular metric that is used to evaluate the results, 3) in sparse representation techniques, L_1 is not a good proxy for the true measure of sparsity, L_0, and 4) the L_epsilon norm may be a better error metric for image reconstruction and compression than the L_2 norm, though the exact psychophysical metric should take into account high order structure in images.

From Regression to Classification in Support Vector Machines

We study the relation between support vector machines (SVMs) for regression (SVMR) and SVM for classification (SVMC). We show that for a given SVMC solution there exists a SVMR solution which is equivalent for a certain choice of the parameters. In particular our result is that for $epsilon$ sufficiently close to one, the optimal hyperplane and threshold for the SVMC problem with regularization parameter C_c are equal to (1-epsilon)^{- 1} times the optimal hyperplane and threshold for SVMR with regularization parameter C_r = (1-epsilon)C_c. A direct consequence of this result is that SVMC can be seen as a special case of SVMR.

On the V(subscript gamma) Dimension for Regression in Reproducing Kernel Hilbert Spaces

This paper presents a computation of the $V_gamma$ dimension for regression in bounded subspaces of Reproducing Kernel Hilbert Spaces (RKHS) for the Support Vector Machine (SVM) regression $epsilon$-insensitive loss function, and general $L_p$ loss functions. Finiteness of the RV_gamma$ dimension is shown, which also proves uniform convergence in probability for regression machines in RKHS subspaces that use the $L_epsilon$ or general $L_p$ loss functions. This paper presenta a novel proof of this result also for the case that a bias is added to the functions in the RKHS.

Direction Estimation of Pedestrian from Images

The capability of estimating the walking direction of people would be useful in many applications such as those involving autonomous cars and robots. We introduce an approach for estimating the walking direction of people from images, based on learning the correct classification of a still image by using SVMs. We find that the performance of the system can be improved by classifying each image of a walking sequence and combining the outputs of the classifier. Experiments were performed to evaluate our system and estimate the trade-off between number of images in walking sequences and performance.

Selecting Relevant Genes with a Spectral Approach

Array technologies have made it possible to record simultaneously the expression pattern of thousands of genes. A fundamental problem in the analysis of gene expression data is the identification of highly relevant genes that either discriminate between phenotypic labels or are important with respect to the cellular process studied in the experiment: for example cell cycle or heat shock in yeast experiments, chemical or genetic perturbations of mammalian cell lines, and genes involved in class discovery for human tumors. In this paper we focus on the task of unsupervised gene selection. The problem of selecting a small subset of genes is particularly challenging as the datasets involved are typically characterized by a very small sample size ?? the order of few tens of tissue samples ??d by a very large feature space as the number of genes tend to be in the high thousands. We propose a model independent approach which scores candidate gene selections using spectral properties of the candidate affinity matrix. The algorithm is very straightforward to implement yet contains a number of remarkable properties which guarantee consistent sparse selections. To illustrate the value of our approach we applied our algorithm on five different datasets. The first consists of time course data from four well studied Hematopoietic cell lines (HL-60, Jurkat, NB4, and U937). The other four datasets include three well studied treatment outcomes (large cell lymphoma, childhood medulloblastomas, breast tumors) and one unpublished dataset (lymph status). We compared our approach both with other unsupervised methods (SOM,PCA,GS) and with supervised methods (SNR,RMB,RFE). The results clearly show that our approach considerably outperforms all the other unsupervised approaches in our study, is competitive with supervised methods and in some case even outperforms supervised approaches.

Support Vector Machines: Training and Applications

The Support Vector Machine (SVM) is a new and very promising classification technique developed by Vapnik and his group at AT&T Bell Labs. This new learning algorithm can be seen as an alternative training technique for Polynomial, Radial Basis Function and Multi-Layer Perceptron classifiers. An interesting property of this approach is that it is an approximate implementation of the Structural Risk Minimization (SRM) induction principle. The derivation of Support Vector Machines, its relationship with SRM, and its geometrical insight, are discussed in this paper. Training a SVM is equivalent to solve a quadratic programming problem with linear and box constraints in a number of variables equal to the number of data points. When the number of data points exceeds few thousands the problem is very challenging, because the quadratic form is completely dense, so the memory needed to store the problem grows with the square of the number of data points. Therefore, training problems arising in some real applications with large data sets are impossible to load into memory, and cannot be solved using standard non-linear constrained optimization algorithms. We present a decomposition algorithm that can be used to train SVM's over large data sets. The main idea behind the decomposition is the iterative solution of sub-problems and the evaluation of, and also establish the stopping criteria for the algorithm. We present previous approaches, as well as results and important details of our implementation of the algorithm using a second-order variant of the Reduced Gradient Method as the solver of the sub-problems. As an application of SVM's, we present preliminary results we obtained applying SVM to the problem of detecting frontal human faces in real images.

A Note on Support Vector Machines Degeneracy

When training Support Vector Machines (SVMs) over non-separable data sets, one sets the threshold $b$ using any dual cost coefficient that is strictly between the bounds of $0$ and $C$. We show that there exist SVM training problems with dual optimal solutions with all coefficients at bounds, but that all such problems are degenerate in the sense that the "optimal separating hyperplane" is given by ${f w} = {f 0}$, and the resulting (degenerate) SVM will classify all future points identically (to the class that supplies more training data). We also derive necessary and sufficient conditions on the input data for this to occur. Finally, we show that an SVM training problem can always be made degenerate by the addition of a single data point belonging to a certain unboundedspolyhedron, which we characterize in terms of its extreme points and rays.

Scene Classification Using a Hybrid Generative/Discriminative Approach

We investigate whether dimensionality reduction using a latent generative model is beneficial for the task of weakly supervised scene classification. In detail, we are given a set of labeled images of scenes (for example, coast, forest, city, river, etc.), and our objective is to classify a new image into one of these categories. Our approach consists of first discovering latent ";topics"; using probabilistic Latent Semantic Analysis (pLSA), a generative model from the statistical text literature here applied to a bag of visual words representation for each image, and subsequently, training a multiway classifier on the topic distribution vector for each image. We compare this approach to that of representing each image by a bag of visual words vector directly and training a multiway classifier on these vectors. To this end, we introduce a novel vocabulary using dense color SIFT descriptors and then investigate the classification performance under changes in the size of the visual vocabulary, the number of latent topics learned, and the type of discriminative classifier used (k-nearest neighbor or SVM). We achieve superior classification performance to recent publications that have used a bag of visual word representation, in all cases, using the authors' own data sets and testing protocols. We also investigate the gain in adding spatial information. We show applications to image retrieval with relevance feedback and to scene classification in videos

Classificació i reconeixement de vídeo

El principal objectiu d’aquest projecte és aconseguir classificar diferents vídeos d’esports segons la seva categoria. Els cercadors de text creen un vocabulari segons el significat de les diferents paraules per tal de poder identificar un document. En aquest projecte es va fer el mateix però mitjançant paraules visuals. Per exemple, es van intentar englobar com a una única paraula les diferents rodes que apareixien en els cotxes de rally. A partir de la freqüència amb què apareixien les paraules dels diferents grups dins d’una imatge vàrem crear histogrames de vocabulari que ens permetien tenir una descripció de la imatge. Per classificar un vídeo es van utilitzar els histogrames que descrivien els seus fotogrames. Com que cada histograma es podia considerar un vector de valors enters vàrem optar per utilitzar una màquina classificadora de vectors: una Support vector machine o SVM

Localización de faltas en sistemas de distribución de energía eléctrica usando métodos basados en el modelo y métodos basados en el conocimiento

La calidad de energía eléctrica incluye la calidad del suministro y la calidad de la atención al cliente. La calidad del suministro a su vez se considera que la conforman dos partes, la forma de onda y la continuidad. En esta tesis se aborda la continuidad del suministro a través de la localización de faltas. Este problema se encuentra relativamente resuelto en los sistemas de transmisión, donde por las características homogéneas de la línea, la medición en ambos terminales y la disponibilidad de diversos equipos, se puede localizar el sitio de falta con una precisión relativamente alta. En sistemas de distribución, sin embargo, la localización de faltas es un problema complejo y aún no resuelto. La complejidad es debida principalmente a la presencia de conductores no homogéneos, cargas intermedias, derivaciones laterales y desbalances en el sistema y la carga. Además, normalmente, en estos sistemas sólo se cuenta con medidas en la subestación, y un modelo simplificado del circuito. Los principales esfuerzos en la localización han estado orientados al desarrollo de métodos que utilicen el fundamental de la tensión y de la corriente en la subestación, para estimar la reactancia hasta la falta. Como la obtención de la reactancia permite cuantificar la distancia al sitio de falta a partir del uso del modelo, el Método se considera Basado en el Modelo (MBM). Sin embargo, algunas de sus desventajas están asociadas a la necesidad de un buen modelo del sistema y a la posibilidad de localizar varios sitios donde puede haber ocurrido la falta, esto es, se puede presentar múltiple estimación del sitio de falta. Como aporte, en esta tesis se presenta un análisis y prueba comparativa entre varios de los MBM frecuentemente referenciados. Adicionalmente se complementa la solución con métodos que utilizan otro tipo de información, como la obtenida de las bases históricas de faltas con registros de tensión y corriente medidos en la subestación (no se limita solamente al fundamental). Como herramienta de extracción de información de estos registros, se utilizan y prueban dos técnicas de clasificación (LAMDA y SVM). Éstas relacionan las características obtenidas de la señal, con la zona bajo falta y se denominan en este documento como Métodos de Clasificación Basados en el Conocimiento (MCBC). La información que usan los MCBC se obtiene de los registros de tensión y de corriente medidos en la subestación de distribución, antes, durante y después de la falta. Los registros se procesan para obtener los siguientes descriptores: a) la magnitud de la variación de tensión ( dV ), b) la variación de la magnitud de corriente ( dI ), c) la variación de la potencia ( dS ), d) la reactancia de falta ( Xf ), e) la frecuencia del transitorio ( f ), y f) el valor propio máximo de la matriz de correlación de corrientes (Sv), cada uno de los cuales ha sido seleccionado por facilitar la localización de la falta. A partir de estos descriptores, se proponen diferentes conjuntos de entrenamiento y validación de los MCBC, y mediante una metodología que muestra la posibilidad de hallar relaciones entre estos conjuntos y las zonas en las cuales se presenta la falta, se seleccionan los de mejor comportamiento. Los resultados de aplicación, demuestran que con la combinación de los MCBC con los MBM, se puede reducir el problema de la múltiple estimación del sitio de falta. El MCBC determina la zona de falta, mientras que el MBM encuentra la distancia desde el punto de medida hasta la falta, la integración en un esquema híbrido toma las mejores características de cada método. En este documento, lo que se conoce como híbrido es la combinación de los MBM y los MCBC, de una forma complementaria. Finalmente y para comprobar los aportes de esta tesis, se propone y prueba un esquema de integración híbrida para localización de faltas en dos sistemas de distribución diferentes. Tanto los métodos que usan los parámetros del sistema y se fundamentan en la estimación de la impedancia (MBM), como aquellos que usan como información los descriptores y se fundamentan en técnicas de clasificación (MCBC), muestran su validez para resolver el problema de localización de faltas. Ambas metodologías propuestas tienen ventajas y desventajas, pero según la teoría de integración de métodos presentada, se alcanza una alta complementariedad, que permite la formulación de híbridos que mejoran los resultados, reduciendo o evitando el problema de la múltiple estimación de la falta.

Investigation of support vector machine for the detection of architectural distortion in mammographic images

This paper investigates detection of architectural distortion in mammographic images using support vector machine. Hausdorff dimension is used to characterise the texture feature of mammographic images. Support vector machine, a learning machine based on statistical learning theory, is trained through supervised learning to detect architectural distortion. Compared to the Radial Basis Function neural networks, SVM produced more accurate classification results in distinguishing architectural distortion abnormality from normal breast parenchyma.

Face verification using Gabor wavelets and AdaBoost

This paper presents a new face verification algorithm based on Gabor wavelets and AdaBoost. In the algorithm, faces are represented by Gabor wavelet features generated by Gabor wavelet transform. Gabor wavelets with 5 scales and 8 orientations are chosen to form a family of Gabor wavelets. By convolving face images with these 40 Gabor wavelets, the original images are transformed into magnitude response images of Gabor wavelet features. The AdaBoost algorithm selects a small set of significant features from the pool of the Gabor wavelet features. Each feature is the basis for a weak classifier which is trained with face images taken from the XM2VTS database. The feature with the lowest classification error is selected in each iteration of the AdaBoost operation. We also address issues regarding computational costs in feature selection with AdaBoost. A support vector machine (SVM) is trained with examples of 20 features, and the results have shown a low false positive rate and a low classification error rate in face verification.

Sparse kernel density construction using orthogonal forward regression with leave-one-out test score and local regularization

This paper presents an efficient construction algorithm for obtaining sparse kernel density estimates based on a regression approach that directly optimizes model generalization capability. Computational efficiency of the density construction is ensured using an orthogonal forward regression, and the algorithm incrementally minimizes the leave-one-out test score. A local regularization method is incorporated naturally into the density construction process to further enforce sparsity. An additional advantage of the proposed algorithm is that it is fully automatic and the user is not required to specify any criterion to terminate the density construction procedure. This is in contrast to an existing state-of-art kernel density estimation method using the support vector machine (SVM), where the user is required to specify some critical algorithm parameter. Several examples are included to demonstrate the ability of the proposed algorithm to effectively construct a very sparse kernel density estimate with comparable accuracy to that of the full sample optimized Parzen window density estimate. Our experimental results also demonstrate that the proposed algorithm compares favorably with the SVM method, in terms of both test accuracy and sparsity, for constructing kernel density estimates.

Characterization and classification of tumor lesions using computerized fractal-based texture analysis and support vector machines in digital mammograms

Objective: This paper presents a detailed study of fractal-based methods for texture characterization of mammographic mass lesions and architectural distortion. The purpose of this study is to explore the use of fractal and lacunarity analysis for the characterization and classification of both tumor lesions and normal breast parenchyma in mammography. Materials and methods: We conducted comparative evaluations of five popular fractal dimension estimation methods for the characterization of the texture of mass lesions and architectural distortion. We applied the concept of lacunarity to the description of the spatial distribution of the pixel intensities in mammographic images. These methods were tested with a set of 57 breast masses and 60 normal breast parenchyma (dataset1), and with another set of 19 architectural distortions and 41 normal breast parenchyma (dataset2). Support vector machines (SVM) were used as a pattern classification method for tumor classification. Results: Experimental results showed that the fractal dimension of region of interest (ROIs) depicting mass lesions and architectural distortion was statistically significantly lower than that of normal breast parenchyma for all five methods. Receiver operating characteristic (ROC) analysis showed that fractional Brownian motion (FBM) method generated the highest area under ROC curve (A z = 0.839 for dataset1, 0.828 for dataset2, respectively) among five methods for both datasets. Lacunarity analysis showed that the ROIs depicting mass lesions and architectural distortion had higher lacunarities than those of ROIs depicting normal breast parenchyma. The combination of FBM fractal dimension and lacunarity yielded the highest A z value (0.903 and 0.875, respectively) than those based on single feature alone for both given datasets. The application of the SVM improved the performance of the fractal-based features in differentiating tumor lesions from normal breast parenchyma by generating higher A z value. Conclusion: FBM texture model is the most appropriate model for characterizing mammographic images due to self-affinity assumption of the method being a better approximation. Lacunarity is an effective counterpart measure of the fractal dimension in texture feature extraction in mammographic images. The classification results obtained in this work suggest that the SVM is an effective method with great potential for classification in mammographic image analysis.

Evaluation of satellite-based and model re-analysis rainfall estimates for Uganda

The dependence of much of Africa on rain fed agriculture leads to a high vulnerability to fluctuations in rainfall amount. Hence, accurate monitoring of near-real time rainfall is particularly useful, for example in forewarning possible crop shortfalls in drought-prone areas. Unfortunately, ground based observations are often inadequate. Rainfall estimates from satellite-based algorithms and numerical model outputs can fill this data gap, however rigorous assessment of such estimates is required. In this case, three satellite based products (NOAA-RFE 2.0, GPCP-1DD and TAMSAT) and two numerical model outputs (ERA-40 and ERA-Interim) have been evaluated for Uganda in East Africa using a network of 27 rain gauges. The study focuses on the years 2001 to 2005 and considers the main rainy season (February to June). All data sets were converted to the same temporal and spatial scales. Kriging was used for the spatial interpolation of the gauge data. All three satellite products showed similar characteristics and had a high level of skill that exceeded both model outputs. ERA-Interim had a tendency to overestimate whilst ERA-40 consistently underestimated the Ugandan rainfall.