Beihang University

About: Beihang University is a education organization based out in Beijing, China. It is known for research contribution in the topics: Control theory & Microstructure. The organization has 67002 authors who have published 73507 publications receiving 975691 citations. The organization is also known as: Beijing University of Aeronautics and Astronautics.

MixUp as Locally Linear Out-Of-Manifold Regularization

Abstract: MixUp (Zhang et al. 2017) is a recently proposed dataaugmentation scheme, which linearly interpolates a random pair of training examples and correspondingly the one-hot representations of their labels. Training deep neural networks with such additional data is shown capable of significantly improving the predictive accuracy of the current art. The power of MixUp, however, is primarily established empirically and its working and effectiveness have not been explained in any depth. In this paper, we develop an understanding for MixUp as a form of “out-of-manifold regularization”, which imposes certain “local linearity” constraints on the model’s input space beyond the data manifold. This analysis enables us to identify a limitation of MixUp, which we call “manifold intrusion”. In a nutshell, manifold intrusion in MixUp is a form of under-fitting resulting from conflicts between the synthetic labels of the mixed-up examples and the labels of original training data. Such a phenomenon usually happens when the parameters controlling the generation of mixing policies are not sufficiently fine-tuned on the training data. To address this issue, we propose a novel adaptive version of MixUp, where the mixing policies are automatically learned from the data using an additional network and objective function designed to avoid manifold intrusion. The proposed regularizer, AdaMixUp, is empirically evaluated on several benchmark datasets. Extensive experiments demonstrate that AdaMixUp improves upon MixUp when applied to the current art of deep classification models.

Anti-disturbance control theory for systems with multiple disturbances: a survey.

Abstract: The problem of anti-disturbance control has been an eternal topic along with the development of the control theory. However, most methodologies can only deal with systems subject to a single equivalent disturbance which was merged by various types of uncertainties. In this paper, a review on anti-disturbance control is presented for systems with multiple disturbances. First, the classical control methods are briefly reviewed for disturbance attenuation or rejection problems. Then, recent advances in disturbance observer based control (DOBC) theory are introduced and especially, the composite hierarchical anti-disturbance control (CHADC) is firstly addressed. A comparison of different approaches is briefly carried out. Finally, focuses in the field on the current research are also addressed with emphasis on the practical application of the techniques.

Nodal surface semimetals: Theory and material realization

Abstract: We theoretically study the three-dimensional topological semimetals with nodal surfaces protected by crystalline symmetries. Different from the well-known nodal-point and nodal-line semimetals, in these materials, the conduction and valence bands cross on closed nodal surfaces in the Brillouin zone. We propose different classes of nodal surfaces, both in the absence and in the presence of spin-orbit coupling (SOC). In the absence of SOC, a class of nodal surfaces can be protected by space-time inversion symmetry and sublattice symmetry and characterized by a ${\mathbb{Z}}_{2}$ index, while another class of nodal surfaces are guaranteed by a combination of nonsymmorphic twofold screw-rotational symmetry and time-reversal symmetry. We show that the inclusion of SOC will destroy the former class of nodal surfaces but may preserve the latter provided that the inversion symmetry is broken. We further generalize the result to magnetically ordered systems and show that protected nodal surfaces can also exist in magnetic materials without and with SOC, given that certain magnetic group symmetry requirements are satisfied. Several concrete nodal-surface material examples are predicted via the first-principles calculations. The possibility of multi-nodal-surface materials is discussed.

Blockchain-Based Solutions to Security and Privacy Issues in the Internet of Things

Abstract: IoT is leading a digital revolution in both academia and industry. It brings convenience to people's daily lives; however, the issues of security and privacy of IoT become challenges. Blockchain, a decentralized database based on cryptographic techniques, is promising for IoT security, which may influence a variety of areas including manufacture, finance, and trading. The blockchain framework in an IoT system is an intriguing alternative to the traditional centralized model, which is struggling to meet some specified demands in IoT. In this article, we investigate typical security and privacy issues in IoT and develop a framework to integrate blockchain with IoT, which can provide great assurance for IoT data and various functionalities and desirable scalability including authentication, decentralized payment, and so on. We also suggest some possible solutions to these security and privacy issues in IoT based on blockchain and Ethereum to show how blockchain contributes to IoT.