The technique of digital twin network (DTN) has been considered as a promising technique of network management and control to accommodate disruptive applications for the sixth generation communication (6G) systems. However, it is still challenging for DTN to provide ultimate experience for the emerging services and applications, due to the lack of effective DTN model management strategies. To solve this problem, the model selection and offloading for DTN is studied in this paper. First, a framework of cooperative model selection and offloading for DTN is designed, which can adapt with the limited computation capability of users, and improve the communication efficiency of model offloading. Second, the performance metrics are proposed to effectively evaluate the accuracy loss and the privacy leakage risk of model management for DTN, and tractable expressions are provided for our studied framework. Third, a joint optimization algorithm is designed for model management and transmit power allocation, which can efficiently reduce the accuracy loss and the privacy leakage risk with low processing latency. Finally, the experiment results are provided to show the effectiveness of our introduced performance metrics, and verify the performance gains of proposed optimization algorithm for our studied framework.

Network performance evaluation is of great significance for network optimization and resource management. However, conventional performance evaluation methods are generally conducted on network areas of cellular scales, which makes it impossible to generate network evaluation results with high geographic precision. To solve this problem and inspired by the research field of computer vision, a deep leaning-based scheme for grid-level performance evaluation of 5G network is proposed and training strategy is designed. First, the grid-level and coarse-grid evaluation modules of the scheme are constructed based on deep learning networks. Secondly, to improve the performance of proposed scheme, a conventional evaluation method is exploited to address the lack of data labels and loss function with regularization term is designed. Moreover, distributed model deployment strategies are provided for the proposed deep learning models. Finally, simulation results illustrate the correlation between generated grid-level evaluation results and input data is strong, which demonstrates the high performance of the proposed scheme.

As the world's largest professional network, LinkedIn wants to create economic opportunity for everyone in the global workforce. One of its most critical missions is matching jobs with processionals. Improving job targeting accuracy and hire efficiency align with LinkedIn's Member First Motto. To achieve those goals, we need to understand unstructured job postings with noisy information. We applied deep transfer learning to create domain-specific job understanding models. After this, jobs are represented by professional entities, including titles, skills, companies, and assessment questions. To continuously improve LinkedIn's job understanding ability, we designed an expert feedback loop where we integrated job understanding models into LinkedIn's products to collect job posters' feedback. In this demonstration, we present LinkedIn's job posting flow and demonstrate how the integrated deep job understanding work improves job posters' satisfaction and provides significant metric lifts in LinkedIn's job recommendation system.

Time-domain analysis and frequency-domain analysis are two angles of view for us to study and survey a continuous function. We observe the function approximation problems from the frequency domain. We consider that a single-frequency sine function can be approximated by two Gaussian kernels in one period. According to that, we present that the first maximum amplitudes as well as their frequencies and initial phases can be used to determine the initial number, centers and widths of radial basis function (RBF) kernels. After the initial structure of an RBF network is determined like that, a small number of RBF kernels can be added in order to further improve the local approximation accuracy. The above viewpoint is verified by two approximation examples.