Search

Highlights • Optimal non-parametric tuning rules specific to quadrotor attitude dynamics are designed. • Optimality of the non-parametric tuning rules is proved. • Efficiency of the tuning rules is verified by simulations and experimentally. Abstract Designing high performance controllers for multirotors is a rigorous task that is often solved by trial and error approach. Trial and error tuning usually results in non-optimal controller parameters. Tuning controllers based on the existing quadrotor models would result in poor performance of quadrotors due to simplifications and inaccuracies in the underlying models. In this paper optimal tuning rules for quadrotor attitude dynamics are designed, which guarantees near-optimal performance and robustness. A single in-flight run of the Modified Relay Feedback Test that takes only few seconds with guaranteed stability is enough to have near-optimal tuning of the controller. The designed tuning rule is tested experimentally in-flight on a custom-built quadrotor. The results showed significant advantages in performance and robustness due to the proposed approach. [ABSTRACT FROM AUTHOR]

In this paper, a new type of multilayer rule-based classifier is proposed and applied to image classification problems. The proposed approach is entirely data-driven and fully automatic. It is generic and can be applied to various classification and prediction problems, but in this paper we focus on image processing, in particular. The core of the classifier is a fully interpretable, understandable, self-organized set of IF…THEN… fuzzy rules based on the prototypes autonomously identified by using a one-pass type training process. The classifier can self-evolve and be updated continuously without a full retraining. Due to the prototype-based nature, it is non-parametric; its training process is non-iterative, highly parallelizable and computationally efficient. At the same time, the proposed approach is able to achieve very high classification accuracy on various benchmark datasets surpassing most of the published methods, be comparable with the human abilities. In addition, it can start classification from the first image of each class in the same way as humans do, which makes the proposed classifier suitable for real-time applications. Numerical examples of benchmark image processing demonstrate the merits of the proposed approach. [ABSTRACT FROM AUTHOR]

The grains of food-type soybean cultivars, which are characterized by the absence of lipoxygenases and the presence of high levels of proteins and isoflavones, are regarded as functional foods with high acceptance by consumers. However, few cultivars of food-type soybeans are currently available in the Brazilian market. The aim of this work was to study the adaptability and stability of various genotypes of food-type soybeans and to compare the performance of methods, which are based on analysis of variance, non-parametric, regression, multivariate and mixed models. Ten lines of food-type soybeans obtained from the Breeding Program of Soybeans for Human Consumption of the State University of Londrina (UEL/BPSHC) and two commercial varieties, the foodtype cultivar BRS 257 and the cultivar BMX Potência RR, were evaluated in the counties of Londrina, Guarapuava, Ponta Grossa and Pato Branco, Paraná, Brazil, during the two sowing seasons of the harvest of 2014/2015. The characteristic evaluated was grain yield. The adaptability parameters of Eberhart and Russel and Cruz methods showed high correlations with the Wricke model. The parameters provided by the analyses of Lins and Binns and REML/BLUP showed higher grain yield associations and moderated correlations with the Eskridge parameters. The AMMI offered the possibility of use in conjunction with the other methodologies. When yield, adaptability and stability were considered, the genotypes UEL 110, UEL 122, UEL 121 and UEL 123 demonstrated potential for the development of new cultivars of food-type soybeans in which lipoxygenases are absent. [ABSTRACT FROM AUTHOR]

This short paper presents some personal considerations on the challenge that the outbreak of big data (which I prefer to call complex data) has posed to statistics as a discipline. I also unpretentiously indicate a possible (hopefully winning) strategy to tackle this challenge. [ABSTRACT FROM AUTHOR]

Estimators for differential entropy are proposed. The estimators are based on the second order expansion of the probability mass around the inspection point with respect to the distance from the point. Simple linear regression is utilized to estimate the values of density function and its second derivative at a point. After estimating the values of the probability density function at each of the given sample points, by taking the empirical average of the negative logarithm of the density estimates, two entropy estimators are derived. Other entropy estimators which directly estimate entropy by linear regression, are also proposed. The proposed four estimators are shown to perform well through numerical experiments for various probability distributions. [ABSTRACT FROM AUTHOR]

New Vapnik–Chervonenkis type concentration inequalities are derived for the empirical distribution of an independent random sample. Focus is on the maximal deviation over classes of Borel sets within a low probability region. The constants are explicit, enabling numerical comparisons. [ABSTRACT FROM AUTHOR]

Abstract: Trimming is necessary for the empirical likelihood inference in the change-point problem. This work studies the asymptotic behavior of the trimmed empirical likelihood ratio (ELR) statistic in its full spectrum. Our results show it is comparable with its parametric counterpart. [Copyright &y& Elsevier]

Abstract: Let be a -valued measurable strictly stationary spatial process, where is a semi-metric space. We study a kernel estimator of the conditional mode of the univariate response variable given the functional variable . The main aim of this paper is to prove the almost complete convergence (with rate) of this estimate [Copyright &y& Elsevier]

We derive the tight constant in the multivariate version of the Dvoretzky–Kiefer–Wolfowitz inequality. The inequality is leveraged to construct the first fully non-parametric test for multivariate probability distributions including a simple formula for the test statistic. We also generalize the test under appropriate α -mixing conditions and describe applications of the tests to machine learning and representative sampling. [ABSTRACT FROM AUTHOR]