Gaussian processes:iterative sparse approximations

In recent years there has been an increased interest in applying non-parametric methods to real-world problems. Significant research has been devoted to Gaussian processes (GPs) due to their increased flexibility when compared with parametric models. These methods use Bayesian learning, which generally leads to analytically intractable posteriors. This thesis proposes a two-step solution to construct a probabilistic approximation to the posterior. In the first step we adapt the Bayesian online learning to GPs: the final approximation to the posterior is the result of propagating the first and second moments of intermediate posteriors obtained by combining a new example with the previous approximation. The propagation of em functional forms is solved by showing the existence of a parametrisation to posterior moments that uses combinations of the kernel function at the training points, transforming the Bayesian online learning of functions into a parametric formulation. The drawback is the prohibitive quadratic scaling of the number of parameters with the size of the data, making the method inapplicable to large datasets. The second step solves the problem of the exploding parameter size and makes GPs applicable to arbitrarily large datasets. The approximation is based on a measure of distance between two GPs, the KL-divergence between GPs. This second approximation is with a constrained GP in which only a small subset of the whole training dataset is used to represent the GP. This subset is called the em Basis Vector, or BV set and the resulting GP is a sparse approximation to the true posterior. As this sparsity is based on the KL-minimisation, it is probabilistic and independent of the way the posterior approximation from the first step is obtained. We combine the sparse approximation with an extension to the Bayesian online algorithm that allows multiple iterations for each input and thus approximating a batch solution. The resulting sparse learning algorithm is a generic one: for different problems we only change the likelihood. The algorithm is applied to a variety of problems and we examine its performance both on more classical regression and classification tasks and to the data-assimilation and a simple density estimation problems.

Bi-directional double auction for financial market simulation

Typical Double Auction (DA) models assume that trading agents are one-way traders. With this limitation, they cannot directly reflect the fact individual traders in financial markets (the most popular application of double auction) choose their trading directions dynamically. To address this issue, we introduce the Bi-directional Double Auction (BDA) market which is populated by two-way traders. Based on experiments under both static and dynamic settings, we find that the allocative efficiency of a static continuous BDA market comes from rational selection of trading directions and is negatively related to the intelligence of trading strategies. Moreover, we introduce Kernel trading strategy designed based on probability density estimation for general DA market. Our experiments show it outperforms some intelligent DA market trading strategies. Copyright © 2013, International Foundation for Autonomous Agents and Multiagent Systems (www.ifaamas.org). All rights reserved.

O risco espacial e fatores associados ao edentulismo em idosos em município do Sudeste do Brasil

Objetivou-se identificar fatores associados ao edentulismo e o seu risco espacial em idosos. Foi realizado um estudo transversal em uma amostra de 372 indivíduos de 60 anos e mais, no Município de Botucatu, São Paulo, Brasil, em 2005. Razões de prevalência brutas e ajustadas foram estimadas por meio de regressão de Poisson, com estimativa robusta da variância e procedimentos de modelagem hierárquica. A análise espacial foi realizada por estimativas de densidade de Kernel. A prevalência de edentulismo foi de 63,17%. Os fatores sociodemográficos associados ao edentulismo foram a baixa escolaridade, o aumento do número de pessoas por cômodo, não possuir automóvel e idade mais avançada, presença de comorbidades, ausência de um cirurgião-dentista regular e ter realizado a última consulta há três anos ou mais. A análise espacial mostrou maior risco nas áreas periféricas. Obteve-se uma melhor compreensão da perda dentária entre os idosos, subsidiando o planejamento de ações em saúde coletiva.

Modelling High-Dimensional Data by Mixtures of Factor Analyzers

We focus on mixtures of factor analyzers from the perspective of a method for model-based density estimation from high-dimensional data, and hence for the clustering of such data. This approach enables a normal mixture model to be fitted to a sample of n data points of dimension p, where p is large relative to n. The number of free parameters is controlled through the dimension of the latent factor space. By working in this reduced space, it allows a model for each component-covariance matrix with complexity lying between that of the isotropic and full covariance structure models. We shall illustrate the use of mixtures of factor analyzers in a practical example that considers the clustering of cell lines on the basis of gene expressions from microarray experiments. (C) 2002 Elsevier Science B.V. All rights reserved.

Renewable power forecast to scheduling of thermal units

In this work is discussed the importance of the renewable production forecast in an island environment. A probabilistic forecast based on kernel density estimators is proposed. The aggregation of these forecasts, allows the determination of thermal generation amount needed to schedule and operating a power grid of an island with high penetration of renewable generation. A case study based on electric system of S. Miguel Island is presented. The results show that the forecast techniques are an imperative tool help the grid management.

Spatial distribution of triatomines in domiciles of an urban area of the Brazilian Southeast Region

Reports of triatomine infestation in urban areas have increased. We analysed the spatial distribution of infestation by triatomines in the urban area of Diamantina, in the state of Minas Gerais, Brazil. Triatomines were obtained by community-based entomological surveillance. Spatial patterns of infestation were analysed by Ripley’s K function and Kernel density estimator. Normalised difference vegetation index (NDVI) and land cover derived from satellite imagery were compared between infested and uninfested areas. A total of 140 adults of four species were captured (100 Triatoma vitticeps, 25Panstrongylus geniculatus, 8 Panstrongylus megistus, and 7 Triatoma arthurneivai specimens). In total, 87.9% were captured within domiciles. Infection by trypanosomes was observed in 19.6% of 107 examined insects. The spatial distributions ofT. vitticeps, P. geniculatus, T. arthurneivai, and trypanosome-positive triatomines were clustered, occurring mainly in peripheral areas. NDVI values were statistically higher in areas infested by T. vitticeps and P. geniculatus. Buildings infested by these species were located closer to open fields, whereas infestations of P. megistus andT. arthurneivai were closer to bare soil. Human occupation and modification of natural areas may be involved in triatomine invasion, exposing the population to these vectors.

A generalization of histogram type estimators

We introduce simple nonparametric density estimators that generalize theclassical histogram and frequency polygon. The new estimators are expressed as linear combination of density functions that are piecewisepolynomials, where the coefficients are optimally chosen in order to minimize the integrated square error of the estimator. We establish the asymptotic behaviour of the proposed estimators, and study theirperformance in a simulation study.

Malliavin calculus in finance

This article is an introduction to Malliavin Calculus for practitioners.We treat one specific application to the calculation of greeks in Finance.We consider also the kernel density method to compute greeks and anextension of the Vega index called the local vega index.

The disturbing 'rise' of global income inequality

We use aggregate GDP data and within-country income shares for theperiod 1970-1998 to assign a level of income to each person in theworld. We then estimate the gaussian kernel density function for theworldwide distribution of income. We compute world poverty rates byintegrating the density function below the poverty lines. The $1/daypoverty rate has fallen from 20% to 5% over the last twenty five years.The $2/day rate has fallen from 44% to 18%. There are between 300 and500 million less poor people in 1998 than there were in the 70s.We estimate global income inequality using seven different popularindexes: the Gini coefficient, the variance of log-income, two ofAtkinson s indexes, the Mean Logarithmic Deviation, the Theil indexand the coefficient of variation. All indexes show a reduction in globalincome inequality between 1980 and 1998. We also find that most globaldisparities can be accounted for by across-country, not within-country,inequalities. Within-country disparities have increased slightly duringthe sample period, but not nearly enough to offset the substantialreduction in across-country disparities. The across-country reductionsin inequality are driven mainly, but not fully, by the large growth rateof the incomes of the 1.2 billion Chinese citizens. Unless Africa startsgrowing in the near future, we project that income inequalities willstart rising again. If Africa does not start growing, then China, India,the OECD and the rest of middle-income and rich countries diverge awayfrom it, and global inequality will rise. Thus, the aggregate GDP growthof the African continent should be the priority of anyone concerned withincreasing global income inequality.

Measuring ecological niche overlap from occurrence and spatial environmental data

1. Aim - Concerns over how global change will influence species distributions, in conjunction with increased emphasis on understanding niche dynamics in evolutionary and community contexts, highlight the growing need for robust methods to quantify niche differences between or within taxa. We propose a statistical framework to describe and compare environmental niches from occurrence and spatial environmental data.¦2. Location - Europe, North America, South America¦3. Methods - The framework applies kernel smoothers to densities of species occurrence in gridded environmental space to calculate metrics of niche overlap and test hypotheses regarding niche conservatism. We use this framework and simulated species with predefined distributions and amounts of niche overlap to evaluate several ordination and species distribution modeling techniques for quantifying niche overlap. We illustrate the approach with data on two well-studied invasive species.¦4. Results - We show that niche overlap can be accurately detected with the framework when variables driving the distributions are known. The method is robust to known and previously undocumented biases related to the dependence of species occurrences on the frequency of environmental conditions that occur across geographic space. The use of a kernel smoother makes the process of moving from geographical space to multivariate environmental space independent of both sampling effort and arbitrary choice of resolution in environmental space. However, the use of ordination and species distribution model techniques for selecting, combining and weighting variables on which niche overlap is calculated provide contrasting results.¦5. Main conclusions - The framework meets the increasing need for robust methods to quantify niche differences. It is appropriate to study niche differences between species, subspecies or intraspecific lineages that differ in their geographical distributions. Alternatively, it can be used to measure the degree to which the environmental niche of a species or intraspecific lineage has changed over time.

Hypernormal densities

We propose a new family of density functions that possess both flexibilityand closed form expressions for moments and anti-derivatives, makingthem particularly appealing for applications. We illustrate its usefulnessby applying our new family to obtain density forecasts of U.S. inflation.Our methods generate forecasts that improve on standard methods based on AR-ARCH models relying on normal or Student's t-distributional assumptions.

Changes in the enterocyte cytoskeleton in newborn rats exposed to ethanol in utero

BACKGROUND: Cytoskeletal changes after longterm exposure to ethanol have been described in a number of cell types in adult rat and humans. These changes can play a key part in the impairment of nutrient assimilation and postnatal growth retardation after prenatal damage of the intestinal epithelium produced by ethanol intake. AIMS: To determine, in the newborn rat, which cytoskeletal proteins are affected by longterm ethanol exposure in utero and to what extent. ANIMALS: The offspring of two experimental groups of female Wistar rats: ethanol treated group receiving up to 25% (w/v) of ethanol in the drinking fluid and control group receiving water as drinking fluid. METHODS: Single and double electron microscopy immunolocalisation and label density estimation of cytoskeletal proteins on sections of proximal small intestine incubated with monoclonal antibodies against actin, alpha-tubulin, cytokeratin (polypeptides 1, 5, 6, 7, 8, 10, 11, and 18), and with a polyclonal antibody anti-beta 1,4-galactosyl transferase as trans golgi (TG) or trans golgi network (TGN) marker, or both. SDS-PAGE technique was also performed on cytoskeletal enriched fractions from small intestine. Western blotting analysis was carried out by incubation with the same antibodies used for immunolocalisation. RESULTS: Intestinal epithelium of newborn rats from the ethanol treated group showed an overexpression of cytoskeletal polypeptides ranging from 39 to 54 kDa, affecting actin and some cytokeratins, but not tubulin. Furthermore, a cytokeratin related polypeptide of 28-29 kDa was identified together with an increase in free ubiquitin in the same group. It was noteworthy that actin and cytokeratin were abnormally located in the TG or the TGN, or both. CONCLUSIONS: Longterm exposure to ethanol in utero causes severe dysfunction in the cytoskeleton of the developing intestinal epithelium. Actin and cytokeratins, which are involved in cytoskeleton anchoring to plasma membrane and cell adhesion, are particularly affected, showing overexpression, impaired proteolysis, and mislocalisation.

Changes in the enterocyte cytoskeleton in newborn rats exposed to ethanol in utero

BACKGROUND: Cytoskeletal changes after longterm exposure to ethanol have been described in a number of cell types in adult rat and humans. These changes can play a key part in the impairment of nutrient assimilation and postnatal growth retardation after prenatal damage of the intestinal epithelium produced by ethanol intake. AIMS: To determine, in the newborn rat, which cytoskeletal proteins are affected by longterm ethanol exposure in utero and to what extent. ANIMALS: The offspring of two experimental groups of female Wistar rats: ethanol treated group receiving up to 25% (w/v) of ethanol in the drinking fluid and control group receiving water as drinking fluid. METHODS: Single and double electron microscopy immunolocalisation and label density estimation of cytoskeletal proteins on sections of proximal small intestine incubated with monoclonal antibodies against actin, alpha-tubulin, cytokeratin (polypeptides 1, 5, 6, 7, 8, 10, 11, and 18), and with a polyclonal antibody anti-beta 1,4-galactosyl transferase as trans golgi (TG) or trans golgi network (TGN) marker, or both. SDS-PAGE technique was also performed on cytoskeletal enriched fractions from small intestine. Western blotting analysis was carried out by incubation with the same antibodies used for immunolocalisation. RESULTS: Intestinal epithelium of newborn rats from the ethanol treated group showed an overexpression of cytoskeletal polypeptides ranging from 39 to 54 kDa, affecting actin and some cytokeratins, but not tubulin. Furthermore, a cytokeratin related polypeptide of 28-29 kDa was identified together with an increase in free ubiquitin in the same group. It was noteworthy that actin and cytokeratin were abnormally located in the TG or the TGN, or both. CONCLUSIONS: Longterm exposure to ethanol in utero causes severe dysfunction in the cytoskeleton of the developing intestinal epithelium. Actin and cytokeratins, which are involved in cytoskeleton anchoring to plasma membrane and cell adhesion, are particularly affected, showing overexpression, impaired proteolysis, and mislocalisation.

Analysis, modelling and classification of geospatial data using machine learning

The research considers the problem of spatial data classification using machine learning algorithms: probabilistic neural networks (PNN) and support vector machines (SVM). As a benchmark model simple k-nearest neighbor algorithm is considered. PNN is a neural network reformulation of well known nonparametric principles of probability density modeling using kernel density estimator and Bayesian optimal or maximum a posteriori decision rules. PNN is well suited to problems where not only predictions but also quantification of accuracy and integration of prior information are necessary. An important property of PNN is that they can be easily used in decision support systems dealing with problems of automatic classification. Support vector machine is an implementation of the principles of statistical learning theory for the classification tasks. Recently they were successfully applied for different environmental topics: classification of soil types and hydro-geological units, optimization of monitoring networks, susceptibility mapping of natural hazards. In the present paper both simulated and real data case studies (low and high dimensional) are considered. The main attention is paid to the detection and learning of spatial patterns by the algorithms applied.

Advanced geostatistical and machine-learning models for spatial data analysis of radioactively contaminated regions

Radioactive soil-contamination mapping and risk assessment is a vital issue for decision makers. Traditional approaches for mapping the spatial concentration of radionuclides employ various regression-based models, which usually provide a single-value prediction realization accompanied (in some cases) by estimation error. Such approaches do not provide the capability for rigorous uncertainty quantification or probabilistic mapping. Machine learning is a recent and fast-developing approach based on learning patterns and information from data. Artificial neural networks for prediction mapping have been especially powerful in combination with spatial statistics. A data-driven approach provides the opportunity to integrate additional relevant information about spatial phenomena into a prediction model for more accurate spatial estimates and associated uncertainty. Machine-learning algorithms can also be used for a wider spectrum of problems than before: classification, probability density estimation, and so forth. Stochastic simulations are used to model spatial variability and uncertainty. Unlike regression models, they provide multiple realizations of a particular spatial pattern that allow uncertainty and risk quantification. This paper reviews the most recent methods of spatial data analysis, prediction, and risk mapping, based on machine learning and stochastic simulations in comparison with more traditional regression models. The radioactive fallout from the Chernobyl Nuclear Power Plant accident is used to illustrate the application of the models for prediction and classification problems. This fallout is a unique case study that provides the challenging task of analyzing huge amounts of data ('hard' direct measurements, as well as supplementary information and expert estimates) and solving particular decision-oriented problems.