MELM-GRBF: A modified version of the extreme learning machine for generalized radial basis function neural networks

In this paper, we propose a methodology for training a new model of artificial neural network called the generalized radial basis function (GRBF) neural network. This model is based on generalized Gaussian distribution, which parametrizes the Gaussian distribution by adding a new parameter τ. The generalized radial basis function allows different radial basis functions to be represented by updating the new parameter τ. For example, when GRBF takes a value of τ=2, it represents the standard Gaussian radial basis function. The model parameters are optimized through a modified version of the extreme learning machine (ELM) algorithm. In the methodology proposed (MELM-GRBF), the centers of each GRBF were taken randomly from the patterns of the training set and the radius and τ values were determined analytically, taking into account that the model must fulfil two constraints: locality and coverage. An thorough experimental study is presented to test its overall performance. Fifteen datasets were considered, including binary and multi-class problems, all of them taken from the UCI repository. The MELM-GRBF was compared to ELM with sigmoidal, hard-limit, triangular basis and radial basis functions in the hidden layer and to the ELM-RBF methodology proposed by Huang et al. (2004) [1]. The MELM-GRBF obtained better results in accuracy than the corresponding sigmoidal, hard-limit, triangular basis and radial basis functions for almost all datasets, producing the highest mean accuracy rank when compared with these other basis functions for all datasets.     

Stochastic approximation learning for mixtures of multivariate elliptical distributions

Most of the current approaches to mixture modeling consider mixture components from a few families of probability distributions, in particular from the Gaussian family. The reasons of these preferences can be traced to their training algorithms, typically versions of the Expectation-Maximization (EM) method. The re-estimation equations needed by this method become very complex as the mixture components depart from the simplest cases. Here we propose to use a stochastic approximation method for probabilistic mixture learning. Under this method it is straightforward to train mixtures composed by a wide range of mixture components from different families. Hence, it is a flexible alternative for mixture learning. Experimental results are presented to show the probability density and missing value estimation capabilities of our proposal.     

基于TS201系列DSP的PGA算法实现

相位梯度自聚焦(PGA)算法被广泛用于机载合成孔径雷达(SAR)成像中.由于其过程复杂,数据处理量大,为满足机载SAR实时处理的要求,选用具有强大处理能力的TS201 DSP.为了充分发挥TS201的性能,利用PGA原理和TS201处理器的特点,对算法流程进行优化,并且对最耗时的FFT采用了汇编语言实现,使得运算量和处...     

A numerical modelling for the extraction of decay regions from color images of monuments

An innovative application focused on the segmentation of decay zones from images of stone materials is presented. The adopted numerical approach to extract decay regions from the color images of monuments provides a tool that helps experts analyze degraded regions by contouring them. In this way even if the results of the proposed procedure depend on the evaluation of experts, the approach can be a contribution to improving the efficiency of the boundary detection process. The segmentation is a process that allows an image to be divided into disjoint zones so that partitioned zones contain homogeneous characteristics. The numerical method, used to segment color images, is based on the theory of interface evolution, which is described by the eikonal equation. We adopted the fast marching technique to solve the upwind finite difference approximation of the eikonal equation. The fast marching starts from a seed point in the region of interest and generates a front which evolves according to a specific speed function until the boundary of the region is identified. We describe the segmentation results obtained with two speed functions, attained by the image gradient computation and color information about the object of interest. Moreover, we present the extension of the working modality of the method by introducing the possibility to extract the regions not only in a local way but also in a global way on the entire image. In this case, in order to improve the segmentation efficiency the application of the fast marching technique starts with more seed points defined as seed regions. The study case concerns the impressive remains of the Roman Theatre in the city of Aosta (Italy). In the image segmentation process the color space L^∗a^∗b^∗$L^{\ast }{a}^{\ast }{b}^{\ast }$ is utilized.     

Medial axis computation for planar free-form shapes

We present a simple, efficient, and stable method for computing—with any desired precision—the medial axis of simply connected planar domains. The domain boundaries are assumed to be given as polynomial spline curves. Our approach combines known results from the field of geometric approximation theory with a new algorithm from the field of computational geometry. Challenging steps are (1) the approximation of the boundary spline such that the medial axis is geometrically stable, and (2) the efficient decomposition of the domain into base cases where the medial axis can be computed directly and exactly. We solve these problems via spiral biarc approximation and a randomized divide & conquer algorithm.     

Finding the longest common nonsuperstring in linear time

String inclusion and non-inclusion problems have been vigorously studied in such diverse fields as molecular biology, data compression, and computer security. Among the well-known string inclusion or non-inclusion notions, we are interested in the longest common nonsuperstring. Given a set of strings, the longest common nonsuperstring problem is finding the longest string that is not a superstring of any string in the given set. It is known that the longest common nonsuperstring problem is solvable in polynomial time.In this paper, we propose an efficient algorithm for the longest common nonsuperstring problem. The running time of our algorithm is linear with respect to the sum of the lengths of the strings in the given set, using generalized suffix trees.     

Discrete data clustering using finite mixture models

Finite mixture models have been applied for different computer vision, image processing and pattern recognition tasks. The majority of the work done concerning finite mixture models has focused on mixtures for continuous data. However, many applications involve and generate discrete data for which discrete mixtures are better suited. In this paper, we investigate the problem of discrete data modeling using finite mixture models. We propose a novel, well motivated mixture that we call the multinomial generalized Dirichlet mixture. The novel model is compared with other discrete mixtures. We designed experiments involving spatial color image databases modeling and summarization, and text classification to show the robustness, flexibility and merits of our approach.     

Combining feature spaces for classification

In this paper we offer a variational Bayes approximation to the multinomial probit model for basis expansion and kernel combination. Our model is well-founded within a hierarchical Bayesian framework and is able to instructively combine available sources of information for multinomial classification. The proposed framework enables informative integration of possibly heterogeneous sources in a multitude of ways, from the simple summation of feature expansions to weighted product of kernels, and it is shown to match and in certain cases outperform the well-known ensemble learning approaches of combining individual classifiers. At the same time the approximation reduces considerably the CPU time and resources required with respect to both the ensemble learning methods and the full Markov chain Monte Carlo, Metropolis-Hastings within Gibbs solution of our model. We present our proposed framework together with extensive experimental studies on synthetic and benchmark datasets and also for the first time report a comparison between summation and product of individual kernels as possible different methods for constructing the composite kernel matrix.     

Numerical approximation of incompressible flows with net flux defective boundary conditions by means of penalty techniques

We consider incompressible flow problems with defective boundary conditions prescribing only the net flux on some inflow and outflow sections of the boundary. As a paradigm for such problems, we simply refer to Stokes flow. After a brief review of the problem and of its well posedness, we discretize the corresponding variational formulation by means of finite elements and looking at the boundary conditions as constraints, we exploit a penalty method to account for them. We perform the analysis of the method in terms of consistency, boundedness and stability of the discrete bilinear form and we show that the application of the penalty method does not affect the optimal convergence properties of the finite element discretization. Since the additional terms introduced to account for the defective boundary conditions are non-local, we also analyze the spectral properties of the equivalent algebraic formulation and we exploit the analysis to set up an efficient solution strategy. In contrast to alternative discretization methods based on Lagrange multipliers accounting for the constraints on the boundary, the present scheme is particularly effective because it only mildly affects the structure and the computational cost of the numerical approximation. Indeed, it does not require neither multipliers nor sub-iterations or additional adjoint problems with respect to the reference problem at hand.