This paper targets on the problem of set to set recognition, which learns the metric between two image sets. Images in each set belong to the same identity. Since images in a set can be complementary, they hopefully lead to higher accuracy in practical applications. However, the quality of each sample cannot be guaranteed, and samples with poor quality will hurt the metric. In this paper, the quality aware network (QAN) is proposed to confront this problem, where the quality of each sample can be automatically learned although such information is not explicitly provided in the training stage. The network has two branches, where the first branch extracts appearance feature embedding for each sample and the other branch predicts quality score for each sample. Features and quality scores of all samples in a set are then aggregated to generate the final feature embedding. We show that the two branches can be trained in an end-to-end manner given only the set-level identity annotation. Analysis on gradient spread of this mechanism indicates that the quality learned by the network is beneficial to set-to-set recognition and simplifies the distribution that the network needs to fit. Experiments on both face verification and person re-identification show advantages of the proposed QAN. The source code and network structure can be downloaded at GitHub 1

Oligoethylene glycols and some related alcohols were efficiently tosylated with p-toluenesulfonyl chloride in a tetrahydrofuran–water (1:1) mixture in the presence of excess sodium hydroxide. This method is advantageous over the conventional tosylation in pyridine both regarding the work-up procedure, the yield, and the purity of the product, and may be potentially useful for the tosylation of certain acid-labile alcohols.

In the calculation of dynamic fault trees, the existing state space–based methods, such as Markov chain method, are basically global-state models, which make the solution procedure very complex. Bayesian networks have become a popular tool to build probability models and conduct inference for reliability design and analysis in various industry fields. The “state explosion” problem can be alleviated by Bayesian networks. Furthermore, to obtain sufficient failure data sets in real engineering systems is extremely difficult and thus causes the parametric uncertainty in failure data. To address these issues, a novel dynamic fault tree analysis method based on the continuous-time Bayesian networks under fuzzy numbers is proposed in this article. The probability distributions under fuzzy numbers for the output variable of dynamic logic gates are determined. The calculation of fuzzy failure probability of a system is presented. Finally, an example is given to demonstrate the effectiveness of the proposed method.

Abstract. In this paper, we discuss the numerical approximation of random periodic solutions of stochastic differential equations (SDEs) with multiplicative noise. We prove the existence of the random periodic solution as the limit of the pull-back flow when the starting time tends to −∞ along the multiple integrals of the period. As the random periodic solution is not explicitly constructible, it is useful to study the numerical approximation. We discretise the SDE using the Euler-Maruyama scheme and modified Milstein scheme. Subsequently, we obtain the existence of the random periodic solution as the limit of the pull-back of the discretised SDE. We prove that the latter is an approximated random periodic solution with an error to the exact one at the rate of √ Δt in the mean square sense in Euler-Maruyama method and Δt in the Milstein method. We also obtain the weak convergence result for the approximation of the periodic measure.Mathematics Subject Classification. 37H99, 60H10, 60H35.

In this paper, autonomous motion control approaches to generate the coordinated motion of a dual-arm space robot for target capturing are presented. Two typical cases are studied: (a) The coordinated dual-arm capturing of a moving target when the base is free-floating; (b) one arm is used for target capturing, and the other for keeping the base fixed inertially. Instead of solving all the variables in a unified differential equation, the solution equation of the first case is simplified into two sub-equations and practical methods are used to solve them. Therefore, the computation loads are largely reduced, and feasible trajectories can be determined. For the second case, we propose to deal with the linear and angular momentums of the system separately. The linear momentum conservation equation is used to design the configuration and the mounted pose of a balance arm to keep the inertial position of the base's center of mass, and the angular momentum conservation equation is used to estimate the desired momentum generated by the reaction wheels for maintaining the inertial attitude of the base. Finally, two typical tasks are simulated. Simulation results verify the corresponding approaches.