Social-Inspired Multicast Feature Selections with Mobile Edge Computing

Ru-Jun Wang, Chih-Hang Wang, De-Nian Yang, Guang-Siang Lee, Wen-Tsuen Chen

Abstract: The emergence of AI has shifted the focus of wireless communications towards deep semantic-level fidelity (i.e., semantic communication networks), emphasizing the significance and effectiveness of transmitted data. However, semantic feature selection considering multicast users with social relations for feature sharing has not been explored. In this paper, we formulate a new optimization problem to minimize the total communication, forwarding, and computation costs, with the proof of NP-hardness and inapproximability. We propose a new algorithm, Multicast Semantic Feature Selection (MSFS), with the notions of Cross Task Semantic Indicator, Substituted Subgraph, and Socially Feature Selection Indicator, to select features on different mobile edge computing servers and cluster the users to receive features via multicast. Simulations with real datasets manifest that MSFS can reduce the total cost by more than 50% compared with state-of-the-art algorithms.