Beijing University of Posts and Telecommunications

About: Beijing University of Posts and Telecommunications is a education organization based out in Beijing, Beijing, China. It is known for research contribution in the topics: MIMO & Quality of service. The organization has 39576 authors who have published 41525 publications receiving 403759 citations. The organization is also known as: BUPT.

Privacy-Preserving Asynchronous Federated Learning Mechanism for Edge Network Computing

Abstract: In the traditional cloud architecture, data needs to be uploaded to the cloud for processing, bringing delays in transmission and response. Edge network emerges as the times require. Data processing on the edge nodes can reduce the delay of data transmission and improve the response speed. In recent years, the need for artificial intelligence of edge network has been proposed. However, the data of a single, individual edge node is limited and does not satisfy the conditions of machine learning. Therefore, performing edge network machine learning under the premise of data confidentiality became a research hotspot. This paper proposes a Privacy-Preserving Asynchronous Federated Learning Mechanism for Edge Network Computing (PAFLM), which can allow multiple edge nodes to achieve more efficient federated learning without sharing their private data. Compared with the traditional distributed learning, the proposed method compresses the communications between nodes and parameter server during the training process without affecting the accuracy. Moreover, it allows the node to join or quit in any process of learning, which can be suitable to the scene with highly mobile edge devices.

Simulated annealing algorithm with adaptive neighborhood

Abstract: As we know, simulated annealing algorithm with large neighborhoods has greater probability of arriving at a global optimum than a small one has, if the other conditions, i.e., the initial configuration, initial temperature and temperature decreasing rate, are the same. However, the large neighborhood is not always beneficial, such as when the distance between the global optimum and the current solution is smaller than the step size. Therefore a simulated annealing algorithm with adaptive neighborhood is proposed in this paper. The adaptive non-uniform mutation operation borrowed from evolutionary algorithm is incorporated into simulated annealing for new solution generation. The neighborhood size reduces in probability with the progress of algorithm. It nearly covers the whole search space in the initial stage of algorithm in the sense of probability. The search engine only searches a very local neighborhood at the later stage of algorithm. Why to hybridize non-uniform mutation with simulated annealing is also analyzed and demonstrated. The numerical experiments show that the hybridization can greatly enhance the performance and the reliability of simulated annealing algorithm. Further experiments are made for benchmarks with expanding search domains. Satisfiable results are obtained again even if the variable bounds are enlarged to 1000 times. Theoretical analysis and simulation experiments illustrate the consistent excellent performance and the possible application of nonu-SA algorithm.

β-NiS modified CdS nanowires for photocatalytic H2 evolution with exceptionally high efficiency.

Abstract: Co-catalysis is regarded as a promising strategy to improve the hydrogen evolution performance of semiconductor-based photocatalysts. But developing a simple and effective technique to achieve the optimal synergy between co-catalysts and host photocatalysts has been a great challenge. Herein, hybrid photocatalysts consisting of β-NiS modified CdS nanowires (NiS/CdS NWs) have been synthesized via a simple and green hydrothermal route using CdS NWs as the template from thiourea and nickel acetate in the presence of sodium hypophosphite. As a result, a metal Ni intermediate was formed via an electroless plating process assisted by H2PO2−, which facilitated the growth of highly conducting flake-like β-NiS nanostructures onto the surface of the CdS NWs. With the optimal loading amount of NiS, the obtained NiS/CdS NWs present a record-high photocatalytic activity for H2 evolution in lactic acid aqueous solutions under visible light irradiation. At 25 °C, the rate of H2 evolution was measured as 793.6 μmol h−1 (over a 5 mg photocatalyst sample), which is nearly 250-fold higher than that over pure CdS NWs, and the apparent quantum yield reached an exceptionally high value of 74.1% at 420 nm. The mechanism for the photocatalytic H2 evolution over the present NiS/CdS NWs was also proposed. This strategy would provide new insight into the design and development of high-performance heterostructured photocatalysts.

End-to-End Low Cost Compressive Spectral Imaging with Spatial-Spectral Self-Attention

Abstract: Coded aperture snapshot spectral imaging (CASSI) is an effective tool to capture real-world 3D hyperspectral images. While a number of existing work has been conducted for hardware and algorithm design, we make a step towards the low-cost solution that enjoys video-rate high-quality reconstruction. To make solid progress on this challenging yet under-investigated task, we reproduce a stable single disperser (SD) CASSI system to gather large-scale real-world CASSI data and propose a novel deep convolutional network to carry out the real-time reconstruction by using self-attention. In order to jointly capture the self-attention across different dimensions in hyperspectral images (i.e., channel-wise spectral correlation and non-local spatial regions), we propose Spatial-Spectral Self-Attention (TSA) to process each dimension sequentially, yet in an order-independent manner. We employ TSA in an encoder-decoder network, dubbed TSA-Net, to reconstruct the desired 3D cube. Furthermore, we investigate how noise affects the results and propose to add shot noise in model training, which improves the real data results significantly. We hope our large-scale CASSI data serve as a benchmark in future research and our TSA model as a baseline in deep learning based reconstruction algorithms. Our code and data are available at https://github.com/mengziyi64/TSA-Net.