We compare two regularization methods which can be used to improve the generalization capabilities of Gaussian mixture density estimates. The rst method consists of deening a Bayesian prior distribution on the parameter space. We derive EM (Expectation Maximization) update rules which maximize the a posterior parameter probability in contrast to the usual EM rules for Gaussian mixtures which ma...

The non-linear inverse problem is formulated in a Bayesian framework. The multivariate normal distribution is assumed in both the noise and prior distributions. However, only the structures of the covariance matrices have to be specified, estimation of the variance levels is included in the inversion procedure. The maximum a posteriori approximation is derived, and the final result is a weighte...

‎in this paper‎, ‎we consider an inverse boundary value problem for two-dimensional heat equation in an annular domain‎. ‎this problem consists of determining the temperature on the interior boundary curve from the cauchy data (boundary temperature and heat flux) on the exterior boundary curve‎. ‎to this end‎, ‎the boundary integral equation method is used‎. ‎since the resulting system of linea...

We study group variable selection on multivariate regression model. Group variable selection is selecting the non-zero rows of coefficient matrix, since there are multiple response variables and thus if one predictor is irrelevant to estimation then the corresponding row must be zero. In a high dimensional setup, shrinkage estimation methods are applicable and guarantee smaller MSE than OLS acc...

Volker Tresp Siemens AG Central Research 81730 Munchen, Germany Volker. [email protected] We compare two regularization methods which can be used to improve the generalization capabilities of Gaussian mixture density estimates. The first method uses a Bayesian prior on the parameter space. We derive EM (Expectation Maximization) update rules which maximize the a posterior parameter probabili...

In this paper, a novel algorithm for image restoration is proposed based on constrained total least-squares (CTLS) estimation, that is, adaptively regularized CTLS (ARCTLS). It is well known that in the regularized CTLS (RCTLS) method, selecting a proper regularization parameter is very difficult. For solving this problem, we take the first-order partial derivative of the classic equation of RC...

Cross-validation (CV) is often used to select the regularization parameter in high dimensional problems. However, when applied to the sparse modeling method Lasso, CV leads to models that are unstable in high-dimensions, and consequently not suited for reliable interpretation. In this paper, we propose a model-free criterion ESCV based on a new estimation stability (ES) metric and CV . Our prop...

this paper presents results of applying a new approach on 2d inversion of magnetotelluric (mt) data in order to enhance resolution and stability of the inversion results. due to non-linearity and limited coverage of data acquisition in an mt field campaign, minimizing the error by linearization of the problem in least squares inversion usually leads to an ill-posed problem. in general, an inver...

This paper is concerned with estimating the solutions of numerically ill-posed least squares problems through Tikhonov regularization. Given a priori estimates on the covariance structure of errors in the measurement data b, and a suitable statistically-chosen σ, the Tikhonov regularized least squares functional J(σ) = ‖Ax − b‖2Wb + 1/σ 2‖D(x − x0)‖2, evaluated at its minimizer x(σ), approximat...