The paper investigates generalized linear models (GLM's) with binary responses such as the logistic, probit, log-log, complementary log-log, scobit and power logit models. It introduces a median estimator of the underlying structural parameters of these models based on statistically smoothed binary responses. Consistency and asymptotic normality of this estimator are proved. Examples of derivation of the asymptotic covariance matrix under the above mentioned models are presented. Finally some comments concerning a method called enhancement and robustness of median estimator are given and results of simulation experiment comparing behavior of median estimator with other robust estimators for GLM's known from the literature are reported.
In this paper, a robust neural network control scheme for the switching dynamical model of the robotic manipulators has been addressed. Radial basis function (RBF) neural networks are employed to approximate unknown functions of robotic manipulators and a compensation controller is designed to enhance system robustness. The weight update law of the robotic manipulator is based on switched multiple Lyapunov function method and the periodically switching law which is suitable for practical implementation is constructed. The proposed control scheme can guarantee that the resulting closed-loop switched system is asymptotically Lyapunov stable and the tracking error performance of the control system is well reached. Finally, a simulation example of two-link robotic manipulators is shown to illustrate the effectiveness of the proposed control method.
The article describes a neural network-based articulatory feature (AF) estimation for the Czech speech. First, the relationship between AFs and a Czech phone inventory is defined, and then the estimation based on the MLP neural networks is done. The usage of several speech representations on the input of the MLP classifiers is proposed with the purpose to obtain a robust AF estimation. The realized experiments have proved that an ANN- based AF estimation works very reliably especially in a low noise environment. Moreover, in case the number of neurons in a hidden layer is increased and if the temporal context DCT-TRAP features are used on the input of the MLP network, the AF classification works accurately also for the signals collected in the environments with a high background noise.
This paper deals with the robust stabilization of a class of nonlinear switched systems with non-vanishing bounded perturbations. The nonlinearities in the systems satisfy a quasi-Lipschitz condition. An observer-based linear-type switching controller with quantized and sampled output signal is considered. Using a dwell-time approach and an extended version of the invariant ellipsoid method (IEM) sufficient conditions for stability in a practical sense are derived. These conditions are represented as Bilinear Matrix Inequalities (BMI's). Finally, two examples are given to verify the efficiency of the proposed method.
This paper presents a novel robust optimal control approach for attitude stabilization of a flexible spacecraft in the presence of external disturbances. An optimal control law is formulated by using concepts of inverse optimal control, proportional-integral-derivative control and a control Lyapunov function. A modified extended state observer is used to compensate for the total disturbances. High-gain and second order sliding mode algorithms are merged to obtain the proposed modified extended state observer. The second method of Lyapunov is used to demonstrate its properties including the convergence rate and ultimate boundedness of the estimation error. The proposed controller can stabilize the attitude control system and minimize a cost functional. Moreover, this controller achieves robustness against bounded external disturbances and the disturbances caused by the elastic vibration of flexible appendages. Numerical simulations are provided to demonstrate the performance of the developed controller.
In this paper, we consider and introduce methods for robust principal component analysis (PCA), including also cases where there are missing values in the data. PCA is a widely applied standard statistical method for data preprocessing, compression, and analysis. It is based on the second-order statistics of the data and is optimal for Gaussian data, but it is often applied to data sets having unknown or other types of probability distributions. PCA can be derived from minimization of the mean-square representation error or maximization of variances under orthonormality constraints. However, these quadratic criteria are sensitive to outliers in the data and long-tailed distributions, which may considerably degrade the results given by PCA. We introduce robust methods for estimation of both the PCA eigenvectors directly or the PCA subspace spanned by them. Experimental results show that our methods provide often better results than standard PCA when outliers are present in the data. Furthermore, we extend our methods to incomplete data with missing values. The problems arising in such cases have several features typical for nonlinear models.
The paper deals with the problem of obtaining a robust PI-D controller design procedure for linear time invariant descriptor uncertain polytopic systems using the regional pole placement and/or H2 criterion approach in the form of a quadratic cost function with the state, derivative state and plant input (QSR). In the frame of Lyapunov Linear Matrix Inequality (LMI) regional pole placement approach and/or H2 quadratic cost function based on Bellman-Lyapunov equation, the designed novel design procedure guarantees the robust properties of closed-loop system with parameter dependent quadratic stability/quadratic stability. In the obtained design procedure the designer could use controller with different structures such as P, PI, PID, PI-D. For the PI-D's D-part of controller feedback the designer could choose any available output/state derivative variables of descriptor systems. Obtained design procedure is in the form of Bilinear Matrix Inequality (BMI). The effectiveness of the obtained results is demonstrated on two examples.
The robust recursive algorithm for the parameter estimation and the volatility prediction in GARCH models is suggested. It seems to be useful for various financial time series, in particular for (high-frequency) log returns contaminated by additive outliers. The proposed procedure can be effective in the risk control and regulation when the prediction of volatility is the main concern since it is capable to distinguish and correct outlaid bursts of volatility. This conclusion is demonstrated by simulations and real data examples presented in the paper.
In this paper, a robust sampled-data observer is proposed for Lipschitz nonlinear systems. Under the minimum-phase condition, it is shown that there always exists a sampling period such that the estimation errors converge to zero for whatever large Lipschitz constant. The optimal sampling period can also be achieved by solving an optimal problem based on linear matrix inequalities (LMIs). The design methods are extended to Lipschitz nonlinear systems with large external disturbances as well. In such a case, the estimation errors converge to a small region of the origin. The size of the region can be small enough by selecting a proper parameter. Compared with the existing results, the design parameters can be easily obtained by solving LMIs.
In this paper we develop the theory of spring balance weighing designs with non-positive correlated errors for that the lower bound of the variance of estimated total weight is attained.