Search

A robust correlation estimator based on the spatial sign covariance matrix (SSCM) is proposed. We derive its asymptotic distribution and influence function at elliptical distributions. Finite sample and robustness properties are studied and compared to other robust correlation estimators by means of numerical simulations. [ABSTRACT FROM AUTHOR]

Abstract: Hierarchical multi-label classification is a complex classification task where the classes involved in the problem are hierarchically structured and each example may simultaneously belong to more than one class in each hierarchical level. In this paper, we extend our previous works, where we investigated a new local-based classification method that incrementally trains a multi-layer perceptron for each level of the classification hierarchy. Predictions made by a neural network in a given level are used as inputs to the neural network responsible for the prediction in the next level. We compare the proposed method with one state-of-the-art decision-tree induction method and two decision-tree induction methods, using several hierarchical multi-label classification datasets. We perform a thorough experimental analysis, showing that our method obtains competitive results to a robust global method regarding both precision and recall evaluation measures. [Copyright &y& Elsevier]

Abstract: This paper investigates the effects of the presence and magnitude of sensor noise on operators' performance and control stability when they use an Ecological Interface Design (EID) interface and a non-EID interface. Sensor noise was gradually increased in selected low-level physical sensors of DURESS III, a representative thermal-hydraulic process simulation. There are two important findings. First, participants in the EID condition achieved target goals significantly faster across all magnitudes of sensor noise. Second, participants in the EID condition exhibited more stable control; experiencing fewer and shorter oscillations around the target goals. This is the first study to empirically investigate the impact of the presence and magnitude of sensor noise on the robustness and effectiveness of an EID interface. These findings are important if EID is to be applied in industrial settings. [Copyright &y& Elsevier]

Abstract: In this work, we cast Theodorsen''s unsteady aerodynamic model into a general form that allows for the introduction of empirically determined quasi-steady and added-mass coefficients as well as an empirical Theodorsen function. An empirically determined Theodorsen model is constructed using data from direct numerical simulations of a flat plate pitching at low Reynolds number, . Next, we develop low-dimensional, state-space realizations that are useful for either the classical Theodorsen lift model or the empirical model. The resulting model is parameterized by pitch-axis location and has physically meaningful states that isolate the effect of added-mass and quasi-steady forces, as well as the effect of the wake. A low-order approximation of Theodorsen''s function is developed based on balanced truncation of a model fit to the analytical frequency response, and it is shown that this approximation outperforms other models from the literature. We demonstrate the utility of these state-space lift models by constructing a robust controller that tracks a reference lift coefficient by varying pitch angle while rejecting gust disturbances. [Copyright &y& Elsevier]

Abstract: Motivating peers to contribute services is critical to the success of peer-to-peer (P2P) systems. Incentive protocols use reciprocity to enforce contributions. Indirect reciprocity schemes are more efficient than direct reciprocity schemes for large-scale P2P systems under high churn rate. In this paper, we propose an indirect reciprocity scheme, called FairTrade, in which peers issue personal currencies to trade services in a P2P system. Personal currency enables indirect reciprocity without relying on any central banks or authorities. It wins extra robustness over global currency as well as much improved trading flexibility and efficiency over direct reciprocity schemes. The acceptance degree of a personal currency depends on the issuer’s service capability and reliance. Peer credit limit is introduced to represent the amount of personal currency that will be accepted by other peers. Every peer as a creditor applies a Bayesian network model to setting peer credit limit for a trading partner peer as a creditee. The Bayesian network model learns the creditee’s capability and reliability and anticipates the associated profits and risks for credit setting. Using simulations on a file-sharing P2P system, we demonstrate that FairTrade achieves success rate of download requests without malicious peers, and maintains over success rate even with malicious nodes. The system warms up quickly and does not assume any altruistic service or other kind of help. On average, the system traffic stabilizes before peers issue their second download requests. All these good performances are achieved with extremely low trading overhead, which takes up less than of the total traffic. [Copyright &y& Elsevier]