Improved Matrix-based Attribute Reduction Algorithm Based on Minimal Elements for Mobile Edge Computing

With the development of Internet of Thlngs(IoT) technology, the Mobile Edge Computing (MEC) environment plays an important role in different scenarios. This paper focus on the "last-mile" delivery scenario in logistics. In this scenario, due to the limited computation ability of the unmanned aerial vehicle (UAV), high real-time and computational resource-intensive tasks such as obstacle avoidance, route planning, and face recognition are require the assistance by MEC computing resources. However, due to the limited computing power of edge servers, the computing tasks of neural networks has very high delay. It cause the deadline of neural networks computation task may not be satisfied. Besides with the users and environment changed continuously, the neural nework take a long time to update which may affect the work of UAV. Therefore, It is an critical problem to accelerate the calculation and update of neural network on edge server. This paper proposes an attribute reduction algorithm based on rough set, which can reduce redundant attributes and compress the dimensionality of data. The experimental results show that can accelerate the calculation and update of neural network significantly and save the energy consumption.