Displaying 101 – 120 of 838

Showing per page

A Generalized Model of PAC Learning and its Applicability

Thomas Brodag, Steffen Herbold, Stephan Waack (2014)

RAIRO - Theoretical Informatics and Applications - Informatique Théorique et Applications

We combine a new data model, where the random classification is subjected to rather weak restrictions which in turn are based on the Mammen−Tsybakov [E. Mammen and A.B. Tsybakov, Ann. Statis. 27 (1999) 1808–1829; A.B. Tsybakov, Ann. Statis. 32 (2004) 135–166.] small margin conditions, and the statistical query (SQ) model due to Kearns [M.J. Kearns, J. ACM 45 (1998) 983–1006] to what we refer to as PAC + SQ model. We generalize the class conditional constant noise (CCCN) model introduced by Decatur...

A geometric approach to correlation inequalities in the plane

A. Figalli, F. Maggi, A. Pratelli (2014)

Annales de l'I.H.P. Probabilités et statistiques

By elementary geometric arguments, correlation inequalities for radially symmetric probability measures are proved in the plane. Precisely, it is shown that the correlation ratio for pairs of width-decreasing sets is minimized within the class of infinite strips. Since open convex sets which are symmetric with respect to the origin turn out to be width-decreasing sets, Pitt’s Gaussian correlation inequality (the two-dimensional case of the long-standing Gaussian correlation conjecture) is derived...

A geometry on the space of probabilities (II). Projective spaces and exponential families.

Henryk Gzyl, Lázaro Recht (2006)

Revista Matemática Iberoamericana

In this note we continue a theme taken up in part I, see [Gzyl and Recht: The geometry on the class of probabilities (I). The finite dimensional case. Rev. Mat. Iberoamericana 22 (2006), 545-558], namely to provide a geometric interpretation of exponential families as end points of geodesics of a non-metric connection in a function space. For that we characterize the space of probability densities as a projective space in the class of strictly positive functions, and these will be regarded as a...

A Global Approach to the Comparison of Clustering Results

Osvaldo Silva, Helena Bacelar-Nicolau, Fernando C. Nicolau (2012)

Biometrical Letters

The discovery of knowledge in the case of Hierarchical Cluster Analysis (HCA) depends on many factors, such as the clustering algorithms applied and the strategies developed in the initial stage of Cluster Analysis. We present a global approach for evaluating the quality of clustering results and making a comparison among different clustering algorithms using the relevant information available (e.g. the stability, isolation and homogeneity of the clusters). In addition, we present a visual method...

A Gram-Schmidt orthogonalizing process of design matrices in linear models as an estimating procedure of covariance components.

Gabriela Beganu (2005)

RACSAM

Se considera un modelo lineal mixto multivariante equilibrado sin interacción para el que las matrices de las formas cuadráticas necesarias para estimar la covarianza de las componentes se expresan mediante operadores lineales en espacios con producto interior de dimensión finita. El propósito de este artículo es demostrar que las formas cuadráticas obtenidas por el proceso de ortogonalización de Gram-Schmidt de las matrices de diseño son combinaciones lineales de las formas cuadráticas derivadas...

A graph-based estimator of the number of clusters

Gérard Biau, Benoît Cadre, Bruno Pelletier (2007)

ESAIM: Probability and Statistics

Assessing the number of clusters of a statistical population is one of the essential issues of unsupervised learning. Given n independent observations X1,...,Xn drawn from an unknown multivariate probability density f, we propose a new approach to estimate the number of connected components, or clusters, of the t-level set ( t ) = { x : f ( x ) t } . The basic idea is to form a rough skeleton of the set ( t ) using any preliminary estimator of f, and to count the number of connected components of the resulting graph. Under...

A homogeneity test of large dimensional covariance matrices under non-normality

M. Rauf Ahmad (2018)

Kybernetika

A test statistic for homogeneity of two or more covariance matrices is presented when the distributions may be non-normal and the dimension may exceed the sample size. Using the Frobenius norm of the difference of null and alternative hypotheses, the statistic is constructed as a linear combination of consistent, location-invariant, estimators of trace functions that constitute the norm. These estimators are defined as U -statistics and the corresponding theory is exploited to derive the normal limit...

A learning algorithm combining functional discriminant coordinates and functional principal components

Tomasz Górecki, Mirosław Krzyśko (2014)

Discussiones Mathematicae Probability and Statistics

A new type of discriminant space for functional data is presented, combining the advantages of a functional discriminant coordinate space and a functional principal component space. In order to provide a comprehensive comparison, we conducted a set of experiments, testing effectiveness on 35 functional data sets (time series). Experiments show that constructed combined space provides a higher quality of classification of LDA method compared with component spaces.

A martingale control variate method for option pricing with stochastic volatility

Jean-Pierre Fouque, Chuan-Hsiang Han (2007)

ESAIM: Probability and Statistics

A generic control variate method is proposed to price options under stochastic volatility models by Monte Carlo simulations. This method provides a constructive way to select control variates which are martingales in order to reduce the variance of unbiased option price estimators. We apply a singular and regular perturbation analysis to characterize the variance reduced by martingale control variates. This variance analysis is done in the regime where time scales of associated driving volatility...

Currently displaying 101 – 120 of 838