SINAPE

"In Neuroscience, to infer how neurons interact to each other is an important problem to understand  how brain works. Currently, there is no technique that makes possible to infer the connectivity of more than hundreds of neurons. As a possible solution to this, we propose as a model of interaction between neurons the Gibbs measure on $mathbb{Z}^d$ having long range interaction and, as the estimation procedure, the $ell_1$ regularized pseudo-maximum likelihood.

Using an extended version of the duality concept between two stochastic processes, we give ergodicity conditions for two states probabilistic cellular automata (PCA) of any dimensions and any radius. Under these assumptions, in the one dimensional case, we study some properties of the unique invariant measure and show that it is shift mixing. Also, the decay of correlation is studied in detail.

A common practice in applied statistics is to determine the sample size under independence. When the available data have an obvious correlation structure the problem is how to determine  the decrease of sample size as a function of correlation. This problem is relevant when a pilot study has been carried out in a certain region and it is of interest to study a regionalized variable in the same area. Recently, some attention has been devoted in the literature to the determination of  geographical sample sizes (Griffith, 2005).

When one wants to perform digital processing of a signal, it must be sampled.  Classically one samples a continuous signal using the Nyquist theorem, that states that the sampling rate of a signal must be at least twice the largest frequency present in it. However, most signals are compressive, that is, their digital version can be alternatively represented with a much smaller number of bits than its original version, with very little information loss. This implies, in general, that they can be represented sparsely using some orthogonal bases.

The Internet has truly revolutionized the way information is gathered and disseminated globally. In the same direction, wireless sensor networks have opened a new lane that promises to extend the Internet's capabilities to include all physical objects and, potentially, all living things.

Nesta palestra serão consideradas abordagens estatísticas em Medicina: baseada em modelos e a algorítmica. A visão de Sackett da estatística baseada em modelos para o clínico.

Exemplos da abordagem algorítmica de Breiman: árvores de classificação, árvores de sobrevida, redes neurais feedfoward, redes neurais probabilísticas, redes de Kohonen, support vector machine, modelos log-lineares e grafos, e visualização de dados.

The ongoing evolution of genomics and bioinformatics has generated massive datasets in enormously large dimensions and in incredible pace. Due to excessive cost of data collection, the sample size is generally disproportionately small, thus leading to the so called high-dimension low sample size (HDLSS) models (K > > n).

In large datasets, classification modeling comprises one of the leading formal tools for supporting the decision making. For instance, in industrial studies a component should be detected for defect. In biomedical studies it is important to determining if a patient is committed with a disease. In financial studies the core objective consists on the generation of a score by means of which potential clients can be listed in order of the probability of default.