计算机科学
计算卸载
计算
GSM演进的增强数据速率
异构网络
边缘计算
分布式计算
计算机网络
人工智能
无线网络
无线
算法
电信
作者
Junhui Zhao,Qiuping Li,Xiaoting Ma,F. Richard Yu
出处
期刊:IEEE Transactions on Network Science and Engineering
[Institute of Electrical and Electronics Engineers]
日期:2023-11-15
卷期号:11 (2): 1872-1884
被引量:4
标识
DOI:10.1109/tnse.2023.3332949
摘要
Drivenby the increasing need of massive data analysis and the rising concern about data privacy, implementing machine learning (ML) at network edge is drawing increasing attention, where local users are expected to process massive raw data without sharing data to a remote central server. However, due to the limited computing power of user equipments, how to deal with the rich data is a critical problem for each user. Based on computation offloading and edge learning, we propose an edge intelligence (EI) learning framework in two-tier heterogeneous networks to alleviate the computing pressure of users. Focusing on the minimum time delay of model training, we analyze the completion time of local learning in parallel manner and obtain the optimal offloading ratio in the proposed EI framework. Aiming at the strict interference constraint of the macrocell base station (MBS), a priority-based power allocation algorithm is designed. The analysis and simulation results verify the proposed algorithm can improve the data transmission rate and reduce the task completion time while satisfying the interference constraints of the MBS and maximum tolerable delay of learning tasks. In addition, the partial computation offloading can effectively improve the learning accuracy within a given learning time budget.
科研通智能强力驱动
Strongly Powered by AbleSci AI