As an emerging distributed machine learning (ML) method, federated learning (FL) can protect data privacy through collaborative learning of artificial intelligence (AI) models across a large number of devices. However, inefficiency and vulnerability to poisoning attacks have slowed FL performance. Therefore, a blockchain-based asynchronous federated learning (BAFL) framework is proposed to ensure the security and efficiency required by FL. The blockchain ensures that the model data cannot be tampered with while asynchronous learning speeds up global aggregation. A novel entropy weight method is used to evaluate the participating rank and proportion of the local model trained in BAFL of the devices. The energy consumption and local model update efficiency are balanced by adjusting the local training and communication delay and optimizing the block generation rate. The extensive evaluation results show that the proposed BAFL framework has higher efficiency and higher performance for preventing poisoning attacks than other distributed ML methods.

This work reports a water-stable Eu ^III -based MOF as the first MOF sensor for detecting nitroaromatic compounds discriminatively by ratiometric methods.

A self-healing conductive G-PyB/KCl hydrogel exhibited rapid room-temperature phase-selective gelation capacity and ultrasensitive detection of alpha-fetoprotein.

The vision of next-generation TV is to support media services to achieve sharing of cross-domain experience, and the eMBB scenario of the 5G network is one of its important driving forces. Considering the bandwidth and resource requirements of different services, such as unicast and multicast services of multimedia TV broadcasting, rationally allocating resources while providing high-quality services and realizing green energy savings of base stations is one of the challenges. This paper is aimed at the resource allocation for TV multimedia service in the 5G wireless cloud network (C-RAN) scenario, which can support unicast services for cellular users and multicast services for broadcast services simultaneously, and it proposes the corresponding slice resources allocation architecture based on the concept of a self-organizing network. The management architecture first builds the functions and processes of the corresponding autonomous resource management. Based on the multidimensional data, an effective deep learning model named LSTM (long short-term memory) is used to construct the dynamic traffic model of the multicast service in space-time, which provides a basis for further network resource allocation. Based on the prediction results and the condition of satisfying the changing requirements of users, the corresponding optimization model is constructed with the goal of minimizing the energy usage of the RRHs (remote radio heads) and taking the QoS constraints of the users into account. A deep reinforcement learning (DRL) framework combined with a convex optimization method are then used to complete the users' bandwidth and power resource allocation. The experimental results show that the proposed method can not only predict the multicast service requirement accurately but also effectively improve the energy efficiency of the network under targeted QoS requirements along with time variations.