Search

Domain adaptation aims at training accurate classifiers in a target domain by utilizing the data in a different but related source domain, which has made great progress in many aspects. Most of existing methods try to match the moments of data by mapping them to feature space, either first or second moment, so as to match the distribution. Despite their appeal, such models often fail to guarantee the obtained features can be well-classified. In this paper, we propose generalized softmax and smooth regularization to extract features and adapt classifiers simultaneously. Considering label matrix as special features, generalized softmax has more tolerance to the diversity of samples belonging to the same class. Smoothness regularization guarantees stronger robustness between target features and decision boundary. Finally, we evaluate our method on several standard benchmark datasets. Empirical evidence shows that the proposed method is comparable or superior to existing methods, and the same results based on two classification schemes indicate that the smoothness regularization is effective. [ABSTRACT FROM AUTHOR]

A modified adaptive backstepping tracking method is proposed to improve the tracking performance of the magnetic bearing system with nonlinear magnetic toque. For a magnetically suspended momentum wheel, two dimensional gyroscopic torque can be produced when the rotor shaft is actively deflected by the active magnetic bearing. High precision rapid tracking control of shaft deflection is desiderated to provide high precision and wide bandwidth outputting torque. The nonlinearity of magnetic bearing is analyzed initially, and the stiffness coefficients of magnetic bearing can be treated as bounded continuous functions with respect to deflection angles. A fuzzy function based adaptive law is proposed to estimate the stiffness coefficients. Combining with a modified backstepping method, the proposed control strategy can deal with the nonlinearity efficiently when the shaft deflects rapidly, and its stability is proved by Lyapunov stability theory. To validate the effectiveness of this method, numerous simulations are performed and the results indicate that this method improves the tracking precision when tracking high frequency reference deflection angles. [ABSTRACT FROM AUTHOR]

Scheduling for the Earth observation satellites (EOSs) imaging mission is a complicated combinatorial optimization problem, especially for the agile EOSs (AEOSs). The increasing observation requirements and orbiting satellites have exacerbated the scheduling complexity in recent years. In this paper, the single agile satellite, redundant observation targets scheduling problem is studied. We introduce the theory of complex networks and find similarities between AEOS redundant targets scheduling problem and the node centrality ranking problem. Then we model this problem as a complex network, regarding each node as a possible observation opportunity, and define two factors, node importance factor and target importance factor, to describe the node/target importance. Based on the two factors, we propose a fast approximate scheduling algorithm (FASA) to obtain the effective scheduling results. Simulation results indicate the FASA is quite efficient and with broad suitability. Our work is helpful in the EOSs and AEOSs scheduling problems by using complex network knowledge. [ABSTRACT FROM AUTHOR]

• This paper utilizes and combines the merits of the assumed enhanced strain method and the numerical manifold method. • The influence of soil non-homogeneity on the static evolution of shear band is investigated naturally and mesh independently. • The simple and useful Gauss Seidel Projection Contraction method is invoked to tackle the difficulties induced by non-smooth yield surfaces. The assumed enhanced strain (AES) method has been successfully used to deal with the strain localization or shear band problems. With AES treating the localization deformation as strong discontinuity in displacement field, the structural phenomenon and the associated strength loss manifested in strain localization could be reflected in a way independent of mesh density and alignment. On the other hand, the numerical manifold method (NMM) excels at dealing with non-homogeneity and discontinuity problems because of the two sets of covers. By combining the NMM with the AES method, this paper could investigate the influence of soil non-homogeneity on shear band evolution naturally and mesh independently. The Gauss Seidel projection contraction (GSPC) method is invoked because it is simple and efficient in tackling the integration difficulties induced by cutting the tensile part of the Drucker Prager (DP) model. Finally, this paper studies the influences of erratic boulders and weak soil layer on the evolution of shear band in the slope. [ABSTRACT FROM AUTHOR]