Cloud-edge-terminal collaborative network (CETCN) has become a key enabler of the next generation wireless network. However, due to the privacy concern, terminals and the edge server may not will to leakage individual data to the cloud server. At the same time, limited computing capability of the edge server and terminals leads to long latency. Therefore, in this paper, we utilize a federated deep reinforce learning (DRL) algorithm named federated learning-based Double Deep Q Network (FL-DDQN) algorithm to solve the task offloading problem in CETCN. To be specific, firstly, we model the task offloading issue as the minimization problem of weighted energy consumption and latency of the CETCN with the constraints of subcarriers and maximum task processing latency. Secondly, we propose a DRL algorithm to obtain the suboptimal solution of the optimization problem. Thirdly, aiming to improve data security and reduce the working pressure of the terminals and the edge server, the FL-DDQN algorithm is further utilized, where the DDQN model is trained cooperatively in the terminals and the edge server. Finally the simulation result demonstrate that our proposed method superior to benchmark solutions in terms of total energy consumption and latency.

The rapid growth of data traffic and use cases brings great challenges to the future mobile communications. Fortunately, service-based architecture (SBA) of 5th generation (5G) core network provides a prospective solution. Hence, a service-based radio access network (RAN) is proposed in this paper to deal with the sophisticated RAN environment and diverse application. Firstly, we present the service-based RAN architecture with four layers and three planes. Secondly, with reference to SBA, we decouple the function layer of the control plane after splitting the control plane and user plane of RAN into several independent control network functions (CNF). Thirdly, a conventional RAN reconfiguration scheme is designed to realize the on-demand management and orchestration (MANO) of service-based RAN. Finally, an experimental platform is built and the experimental results show that compared with the traditional RAN, the response time of service-based RAN is reduced obviously.

The judgements in some sports competitions are subjective tasks, especially in competitions with high requirements on skills, which could lead to unfair results. Synchronized diving is such a skillful competition. Using computer to judge competitions automatically or assist referees to judge can effectively decrease the unfair results. This paper presents a framework for analyzing synchronization for synchronized diving videos. The framework is composed of three steps: (1) silhouette extraction, that can effectively get the diverpsilas silhouette; (2) synchronization feature extraction and representation, that can explicitly achieve synchronization features from synchronized diving videos; (3) synchronization evaluation, that can evaluate synchronization with evaluation function which is constructed by statistic learning method with diving video segments come from different games judged by different referees. The experiment shows that the method can accurately and effectively evaluate the synchronization for synchronized diving videos.

Usually, the image search results contain multiple topics on semantic level and even semantically consistent images have diverse appearances on visual level. How to organize the results into semantically and visually consistent clusters becomes a necessary task to facilitate users' navigation. To attack this, HiCluster, an effective method to organize image search results is designed in this paper, which employs both textual and visual analysis. First, we extract some query-related key phrases to enumerate specific semantics of the given query and cluster them into some semantic clusters using K-lines-based clustering algorithm. Second, the resulting images corresponding to each key phrase are clustered with Bregman Bubble Clustering (BBC) algorithm, which partially groups images in the whole set while discarding some scattered noisy ones. At last, a novel user interface (UI) is designed to provide users with the diverse and helpful information based on the hierarchical clustering structure. Experiments on web images demonstrate the effectiveness and potential of the system.